tccpp.c

   1 /*
   2  *  TCC - Tiny C Compiler
   3  *
   4  *  Copyright (c) 2001-2004 Fabrice Bellard
   5  *
   6  * This library is free software; you can redistribute it and/or
   7  * modify it under the terms of the GNU Lesser General Public
   8  * License as published by the Free Software Foundation; either
   9  * version 2 of the License, or (at your option) any later version.
  10  *
  11  * This library is distributed in the hope that it will be useful,
  12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  14  * Lesser General Public License for more details.
  15  *
  16  * You should have received a copy of the GNU Lesser General Public
  17  * License along with this library; if not, write to the Free Software
  18  * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  19  */
  20
  21 #include "tcc.h"
  22
  23 /********************************************************/
  24 /* global variables */
  25
  26 ST_DATA int tok_flags;
  27 /* additional informations about token */
  28 #define TOK_FLAG_BOL   0x0001 /* beginning of line before */
  29 #define TOK_FLAG_BOF   0x0002 /* beginning of file before */
  30 #define TOK_FLAG_ENDIF 0x0004 /* a endif was found matching starting #ifdef */
  31 #define TOK_FLAG_EOF   0x0008 /* end of file */
  32
  33 ST_DATA int parse_flags;
  34 #define PARSE_FLAG_PREPROCESS 0x0001 /* activate preprocessing */
  35 #define PARSE_FLAG_TOK_NUM    0x0002 /* return numbers instead of TOK_PPNUM */
  36 #define PARSE_FLAG_LINEFEED   0x0004 /* line feed is returned as a
  37                                         token. line feed is also
  38                                         returned at eof */
  39 #define PARSE_FLAG_ASM_COMMENTS 0x0008 /* '#' can be used for line comment */
  40 #define PARSE_FLAG_SPACES     0x0010 /* next() returns space tokens (for -E) */
  41
  42 ST_DATA struct BufferedFile *file;
  43 ST_DATA int ch, tok;
  44 ST_DATA CValue tokc;
  45 ST_DATA const int *macro_ptr;
  46 ST_DATA CString tokcstr; /* current parsed string, if any */
  47
  48 /* display benchmark infos */
  49 ST_DATA int total_lines;
  50 ST_DATA int total_bytes;
  51 ST_DATA int tok_ident;
  52 ST_DATA TokenSym **table_ident;
  53
  54 /* ------------------------------------------------------------------------- */
  55
  56 static int *macro_ptr_allocated;
  57 static const int *unget_saved_macro_ptr;
  58 static int unget_saved_buffer[TOK_MAX_SIZE + 1];
  59 static int unget_buffer_enabled;
  60 static TokenSym *hash_ident[TOK_HASH_SIZE];
  61 static char token_buf[STRING_MAX_SIZE + 1];
  62 /* true if isid(c) || isnum(c) */
  63 static unsigned char isidnum_table[256-CH_EOF];
  64
  65 static const char tcc_keywords[] =
  66 #define DEF(id, str) str "\0"
  67 #include "tcctok.h"
  68 #undef DEF
  69 ;
  70
  71 /* WARNING: the content of this string encodes token numbers */
  72 static const unsigned char tok_two_chars[] =
  73 /* outdated -- gr
  74     "<=\236>=\235!=\225&&\240||\241++\244--\242==\224<<\1>>\2+=\253"
  75     "-=\255*=\252/=\257%=\245&=\246^=\336|=\374->\313..\250##\266";
  76 */{
  77     '<','=', TOK_LE,
  78     '>','=', TOK_GE,
  79     '!','=', TOK_NE,
  80     '&','&', TOK_LAND,
  81     '|','|', TOK_LOR,
  82     '+','+', TOK_INC,
  83     '-','-', TOK_DEC,
  84     '=','=', TOK_EQ,
  85     '<','<', TOK_SHL,
  86     '>','>', TOK_SAR,
  87     '+','=', TOK_A_ADD,
  88     '-','=', TOK_A_SUB,
  89     '*','=', TOK_A_MUL,
  90     '/','=', TOK_A_DIV,
  91     '%','=', TOK_A_MOD,
  92     '&','=', TOK_A_AND,
  93     '^','=', TOK_A_XOR,
  94     '|','=', TOK_A_OR,
  95     '-','>', TOK_ARROW,
  96     '.','.', 0xa8, // C++ token ?
  97     '#','#', TOK_TWOSHARPS,
  98     0
  99 };
 100
 101 struct macro_level {
 102     struct macro_level *prev;
 103     const int *p;
 104 };
 105
 106 static void next_nomacro_spc(void);
 107 static void macro_subst(
 108     TokenString *tok_str,
 109     Sym **nested_list,
 110     const int *macro_str,
 111     struct macro_level **can_read_stream
 112     );
 113
 114 ST_FUNC void skip(int c)
 115 {
 116     if (tok != c)
 117         tcc_error("'%c' expected (got \"%s\")", c, get_tok_str(tok, &tokc));
 118     next();
 119 }
 120
 121 ST_FUNC void expect(const char *msg)
 122 {
 123     tcc_error("%s expected", msg);
 124 }
 125
 126 /* ------------------------------------------------------------------------- */
 127 /* CString handling */
 128 static void cstr_realloc(CString *cstr, int new_size)
 129 {
 130     int size;
 131     void *data;
 132
 133     size = cstr->size_allocated;
 134     if (size == 0)
 135         size = 8; /* no need to allocate a too small first string */
 136     while (size < new_size)
 137         size = size * 2;
 138     data = tcc_realloc(cstr->data_allocated, size);
 139     cstr->data_allocated = data;
 140     cstr->size_allocated = size;
 141     cstr->data = data;
 142 }
 143
 144 /* add a byte */
 145 ST_FUNC void cstr_ccat(CString *cstr, int ch)
 146 {
 147     int size;
 148     size = cstr->size + 1;
 149     if (size > cstr->size_allocated)
 150         cstr_realloc(cstr, size);
 151     ((unsigned char *)cstr->data)[size - 1] = ch;
 152     cstr->size = size;
 153 }
 154
 155 ST_FUNC void cstr_cat(CString *cstr, const char *str)
 156 {
 157     int c;
 158     for(;;) {
 159         c = *str;
 160         if (c == '\0')
 161             break;
 162         cstr_ccat(cstr, c);
 163         str++;
 164     }
 165 }
 166
 167 /* add a wide char */
 168 ST_FUNC void cstr_wccat(CString *cstr, int ch)
 169 {
 170     int size;
 171     size = cstr->size + sizeof(nwchar_t);
 172     if (size > cstr->size_allocated)
 173         cstr_realloc(cstr, size);
 174     *(nwchar_t *)(((unsigned char *)cstr->data) + size - sizeof(nwchar_t)) = ch;
 175     cstr->size = size;
 176 }
 177
 178 ST_FUNC void cstr_new(CString *cstr)
 179 {
 180     memset(cstr, 0, sizeof(CString));
 181 }
 182
 183 /* free string and reset it to NULL */
 184 ST_FUNC void cstr_free(CString *cstr)
 185 {
 186     tcc_free(cstr->data_allocated);
 187     cstr_new(cstr);
 188 }
 189
 190 /* reset string to empty */
 191 ST_FUNC void cstr_reset(CString *cstr)
 192 {
 193     cstr->size = 0;
 194 }
 195
 196 /* XXX: unicode ? */
 197 static void add_char(CString *cstr, int c)
 198 {
 199     if (c == '\'' || c == '\"' || c == '\\') {
 200         /* XXX: could be more precise if char or string */
 201         cstr_ccat(cstr, '\\');
 202     }
 203     if (c >= 32 && c <= 126) {
 204         cstr_ccat(cstr, c);
 205     } else {
 206         cstr_ccat(cstr, '\\');
 207         if (c == '\n') {
 208             cstr_ccat(cstr, 'n');
 209         } else {
 210             cstr_ccat(cstr, '0' + ((c >> 6) & 7));
 211             cstr_ccat(cstr, '0' + ((c >> 3) & 7));
 212             cstr_ccat(cstr, '0' + (c & 7));
 213         }
 214     }
 215 }
 216
 217 /* ------------------------------------------------------------------------- */
 218 /* allocate a new token */
 219 static TokenSym *tok_alloc_new(TokenSym **pts, const char *str, int len)
 220 {
 221     TokenSym *ts, **ptable;
 222     int i;
 223
 224     if (tok_ident >= SYM_FIRST_ANOM)
 225         tcc_error("memory full (symbols)");
 226
 227     /* expand token table if needed */
 228     i = tok_ident - TOK_IDENT;
 229     if ((i % TOK_ALLOC_INCR) == 0) {
 230         ptable = tcc_realloc(table_ident, (i + TOK_ALLOC_INCR) * sizeof(TokenSym *));
 231         table_ident = ptable;
 232     }
 233
 234     ts = tcc_malloc(sizeof(TokenSym) + len);
 235     table_ident[i] = ts;
 236     ts->tok = tok_ident++;
 237     ts->sym_define = NULL;
 238     ts->sym_label = NULL;
 239     ts->sym_struct = NULL;
 240     ts->sym_identifier = NULL;
 241     ts->len = len;
 242     ts->hash_next = NULL;
 243     memcpy(ts->str, str, len);
 244     ts->str[len] = '\0';
 245     *pts = ts;
 246     return ts;
 247 }
 248
 249 #define TOK_HASH_INIT 1
 250 #define TOK_HASH_FUNC(h, c) ((h) * 263 + (c))
 251
 252 /* find a token and add it if not found */
 253 ST_FUNC TokenSym *tok_alloc(const char *str, int len)
 254 {
 255     TokenSym *ts, **pts;
 256     int i;
 257     unsigned int h;
 258
 259     h = TOK_HASH_INIT;
 260     for(i=0;i<len;i++)
 261         h = TOK_HASH_FUNC(h, ((unsigned char *)str)[i]);
 262     h &= (TOK_HASH_SIZE - 1);
 263
 264     pts = &hash_ident[h];
 265     for(;;) {
 266         ts = *pts;
 267         if (!ts)
 268             break;
 269         if (ts->len == len && !memcmp(ts->str, str, len))
 270             return ts;
 271         pts = &(ts->hash_next);
 272     }
 273     return tok_alloc_new(pts, str, len);
 274 }
 275
 276 /* XXX: buffer overflow */
 277 /* XXX: float tokens */
 278 ST_FUNC char *get_tok_str(int v, CValue *cv)
 279 {
 280     static char buf[STRING_MAX_SIZE + 1];
 281     static CString cstr_buf;
 282     CString *cstr;
 283     char *p;
 284     int i, len;
 285
 286     /* NOTE: to go faster, we give a fixed buffer for small strings */
 287     cstr_reset(&cstr_buf);
 288     cstr_buf.data = buf;
 289     cstr_buf.size_allocated = sizeof(buf);
 290     p = buf;
 291
 292 /* just an explanation, should never happen:
 293     if (v <= TOK_LINENUM && v >= TOK_CINT && cv == NULL)
 294         tcc_error("internal error: get_tok_str"); */
 295
 296     switch(v) {
 297     case TOK_CINT:
 298     case TOK_CUINT:
 299         /* XXX: not quite exact, but only useful for testing */
 300         sprintf(p, "%u", cv->ui);
 301         break;
 302     case TOK_CLLONG:
 303     case TOK_CULLONG:
 304         /* XXX: not quite exact, but only useful for testing  */
 305 #ifdef _WIN32
 306         sprintf(p, "%u", (unsigned)cv->ull);
 307 #else
 308         sprintf(p, "%llu", cv->ull);
 309 #endif
 310         break;
 311     case TOK_LCHAR:
 312         cstr_ccat(&cstr_buf, 'L');
 313     case TOK_CCHAR:
 314         cstr_ccat(&cstr_buf, '\'');
 315         add_char(&cstr_buf, cv->i);
 316         cstr_ccat(&cstr_buf, '\'');
 317         cstr_ccat(&cstr_buf, '\0');
 318         break;
 319     case TOK_PPNUM:
 320         cstr = cv->cstr;
 321         len = cstr->size - 1;
 322         for(i=0;i<len;i++)
 323             add_char(&cstr_buf, ((unsigned char *)cstr->data)[i]);
 324         cstr_ccat(&cstr_buf, '\0');
 325         break;
 326     case TOK_LSTR:
 327         cstr_ccat(&cstr_buf, 'L');
 328     case TOK_STR:
 329         cstr = cv->cstr;
 330         cstr_ccat(&cstr_buf, '\"');
 331         if (v == TOK_STR) {
 332             len = cstr->size - 1;
 333             for(i=0;i<len;i++)
 334                 add_char(&cstr_buf, ((unsigned char *)cstr->data)[i]);
 335         } else {
 336             len = (cstr->size / sizeof(nwchar_t)) - 1;
 337             for(i=0;i<len;i++)
 338                 add_char(&cstr_buf, ((nwchar_t *)cstr->data)[i]);
 339         }
 340         cstr_ccat(&cstr_buf, '\"');
 341         cstr_ccat(&cstr_buf, '\0');
 342         break;
 343
 344     case TOK_CFLOAT:
 345     case TOK_CDOUBLE:
 346     case TOK_CLDOUBLE:
 347     case TOK_LINENUM:
 348         return NULL; /* should not happen */
 349
 350     /* above tokens have value, the ones below don't */
 351
 352     case TOK_LT:
 353         v = '<';
 354         goto addv;
 355     case TOK_GT:
 356         v = '>';
 357         goto addv;
 358     case TOK_DOTS:
 359         return strcpy(p, "...");
 360     case TOK_A_SHL:
 361         return strcpy(p, "<<=");
 362     case TOK_A_SAR:
 363         return strcpy(p, ">>=");
 364     default:
 365         if (v < TOK_IDENT) {
 366             /* search in two bytes table */
 367             const unsigned char *q = tok_two_chars;
 368             while (*q) {
 369                 if (q[2] == v) {
 370                     *p++ = q[0];
 371                     *p++ = q[1];
 372                     *p = '\0';
 373                     return buf;
 374                 }
 375                 q += 3;
 376             }
 377         addv:
 378             *p++ = v;
 379             *p = '\0';
 380         } else if (v < tok_ident) {
 381             return table_ident[v - TOK_IDENT]->str;
 382         } else if (v >= SYM_FIRST_ANOM) {
 383             /* special name for anonymous symbol */
 384             sprintf(p, "L.%u", v - SYM_FIRST_ANOM);
 385         } else {
 386             /* should never happen */
 387             return NULL;
 388         }
 389         break;
 390     }
 391     return cstr_buf.data;
 392 }
 393
 394 /* fill input buffer and peek next char */
 395 static int tcc_peekc_slow(BufferedFile *bf)
 396 {
 397     int len;
 398     /* only tries to read if really end of buffer */
 399     if (bf->buf_ptr >= bf->buf_end) {
 400         if (bf->fd != -1) {
 401 #if defined(PARSE_DEBUG)
 402             len = 8;
 403 #else
 404             len = IO_BUF_SIZE;
 405 #endif
 406             len = read(bf->fd, bf->buffer, len);
 407             if (len < 0)
 408                 len = 0;
 409         } else {
 410             len = 0;
 411         }
 412         total_bytes += len;
 413         bf->buf_ptr = bf->buffer;
 414         bf->buf_end = bf->buffer + len;
 415         *bf->buf_end = CH_EOB;
 416     }
 417     if (bf->buf_ptr < bf->buf_end) {
 418         return bf->buf_ptr[0];
 419     } else {
 420         bf->buf_ptr = bf->buf_end;
 421         return CH_EOF;
 422     }
 423 }
 424
 425 /* return the current character, handling end of block if necessary
 426    (but not stray) */
 427 ST_FUNC int handle_eob(void)
 428 {
 429     return tcc_peekc_slow(file);
 430 }
 431
 432 /* read next char from current input file and handle end of input buffer */
 433 ST_INLN void inp(void)
 434 {
 435     ch = *(++(file->buf_ptr));
 436     /* end of buffer/file handling */
 437     if (ch == CH_EOB)
 438         ch = handle_eob();
 439 }
 440
 441 /* handle '\[\r]\n' */
 442 static int handle_stray_noerror(void)
 443 {
 444     while (ch == '\\') {
 445         inp();
 446         if (ch == '\n') {
 447             file->line_num++;
 448             inp();
 449         } else if (ch == '\r') {
 450             inp();
 451             if (ch != '\n')
 452                 goto fail;
 453             file->line_num++;
 454             inp();
 455         } else {
 456         fail:
 457             return 1;
 458         }
 459     }
 460     return 0;
 461 }
 462
 463 static void handle_stray(void)
 464 {
 465     if (handle_stray_noerror())
 466         tcc_error("stray '\\' in program");
 467 }
 468
 469 /* skip the stray and handle the \\n case. Output an error if
 470    incorrect char after the stray */
 471 static int handle_stray1(uint8_t *p)
 472 {
 473     int c;
 474
 475     if (p >= file->buf_end) {
 476         file->buf_ptr = p;
 477         c = handle_eob();
 478         p = file->buf_ptr;
 479         if (c == '\\')
 480             goto parse_stray;
 481     } else {
 482     parse_stray:
 483         file->buf_ptr = p;
 484         ch = *p;
 485         handle_stray();
 486         p = file->buf_ptr;
 487         c = *p;
 488     }
 489     return c;
 490 }
 491
 492 /* handle just the EOB case, but not stray */
 493 #define PEEKC_EOB(c, p)\
 494 {\
 495     p++;\
 496     c = *p;\
 497     if (c == '\\') {\
 498         file->buf_ptr = p;\
 499         c = handle_eob();\
 500         p = file->buf_ptr;\
 501     }\
 502 }
 503
 504 /* handle the complicated stray case */
 505 #define PEEKC(c, p)\
 506 {\
 507     p++;\
 508     c = *p;\
 509     if (c == '\\') {\
 510         c = handle_stray1(p);\
 511         p = file->buf_ptr;\
 512     }\
 513 }
 514
 515 /* input with '\[\r]\n' handling. Note that this function cannot
 516    handle other characters after '\', so you cannot call it inside
 517    strings or comments */
 518 ST_FUNC void minp(void)
 519 {
 520     inp();
 521     if (ch == '\\')
 522         handle_stray();
 523 }
 524
 525
 526 /* single line C++ comments */
 527 static uint8_t *parse_line_comment(uint8_t *p)
 528 {
 529     int c;
 530
 531     p++;
 532     for(;;) {
 533         c = *p;
 534     redo:
 535         if (c == '\n' || c == CH_EOF) {
 536             break;
 537         } else if (c == '\\') {
 538             file->buf_ptr = p;
 539             c = handle_eob();
 540             p = file->buf_ptr;
 541             if (c == '\\') {
 542                 PEEKC_EOB(c, p);
 543                 if (c == '\n') {
 544                     file->line_num++;
 545                     PEEKC_EOB(c, p);
 546                 } else if (c == '\r') {
 547                     PEEKC_EOB(c, p);
 548                     if (c == '\n') {
 549                         file->line_num++;
 550                         PEEKC_EOB(c, p);
 551                     }
 552                 }
 553             } else {
 554                 goto redo;
 555             }
 556         } else {
 557             p++;
 558         }
 559     }
 560     return p;
 561 }
 562
 563 /* C comments */
 564 ST_FUNC uint8_t *parse_comment(uint8_t *p)
 565 {
 566     int c;
 567
 568     p++;
 569     for(;;) {
 570         /* fast skip loop */
 571         for(;;) {
 572             c = *p;
 573             if (c == '\n' || c == '*' || c == '\\')
 574                 break;
 575             p++;
 576             c = *p;
 577             if (c == '\n' || c == '*' || c == '\\')
 578                 break;
 579             p++;
 580         }
 581         /* now we can handle all the cases */
 582         if (c == '\n') {
 583             file->line_num++;
 584             p++;
 585         } else if (c == '*') {
 586             p++;
 587             for(;;) {
 588                 c = *p;
 589                 if (c == '*') {
 590                     p++;
 591                 } else if (c == '/') {
 592                     goto end_of_comment;
 593                 } else if (c == '\\') {
 594                     file->buf_ptr = p;
 595                     c = handle_eob();
 596                     p = file->buf_ptr;
 597                     if (c == '\\') {
 598                         /* skip '\[\r]\n', otherwise just skip the stray */
 599                         while (c == '\\') {
 600                             PEEKC_EOB(c, p);
 601                             if (c == '\n') {
 602                                 file->line_num++;
 603                                 PEEKC_EOB(c, p);
 604                             } else if (c == '\r') {
 605                                 PEEKC_EOB(c, p);
 606                                 if (c == '\n') {
 607                                     file->line_num++;
 608                                     PEEKC_EOB(c, p);
 609                                 }
 610                             } else {
 611                                 goto after_star;
 612                             }
 613                         }
 614                     }
 615                 } else {
 616                     break;
 617                 }
 618             }
 619         after_star: ;
 620         } else {
 621             /* stray, eob or eof */
 622             file->buf_ptr = p;
 623             c = handle_eob();
 624             p = file->buf_ptr;
 625             if (c == CH_EOF) {
 626                 tcc_error("unexpected end of file in comment");
 627             } else if (c == '\\') {
 628                 p++;
 629             }
 630         }
 631     }
 632  end_of_comment:
 633     p++;
 634     return p;
 635 }
 636
 637 #define cinp minp
 638
 639 static inline void skip_spaces(void)
 640 {
 641     while (is_space(ch))
 642         cinp();
 643 }
 644
 645 static inline int check_space(int t, int *spc)
 646 {
 647     if (is_space(t)) {
 648         if (*spc)
 649             return 1;
 650         *spc = 1;
 651     } else
 652         *spc = 0;
 653     return 0;
 654 }
 655
 656 /* parse a string without interpreting escapes */
 657 static uint8_t *parse_pp_string(uint8_t *p,
 658                                 int sep, CString *str)
 659 {
 660     int c;
 661     p++;
 662     for(;;) {
 663         c = *p;
 664         if (c == sep) {
 665             break;
 666         } else if (c == '\\') {
 667             file->buf_ptr = p;
 668             c = handle_eob();
 669             p = file->buf_ptr;
 670             if (c == CH_EOF) {
 671             unterminated_string:
 672                 /* XXX: indicate line number of start of string */
 673                 tcc_error("missing terminating %c character", sep);
 674             } else if (c == '\\') {
 675                 /* escape : just skip \[\r]\n */
 676                 PEEKC_EOB(c, p);
 677                 if (c == '\n') {
 678                     file->line_num++;
 679                     p++;
 680                 } else if (c == '\r') {
 681                     PEEKC_EOB(c, p);
 682                     if (c != '\n')
 683                         expect("'\n' after '\r'");
 684                     file->line_num++;
 685                     p++;
 686                 } else if (c == CH_EOF) {
 687                     goto unterminated_string;
 688                 } else {
 689                     if (str) {
 690                         cstr_ccat(str, '\\');
 691                         cstr_ccat(str, c);
 692                     }
 693                     p++;
 694                 }
 695             }
 696         } else if (c == '\n') {
 697             file->line_num++;
 698             goto add_char;
 699         } else if (c == '\r') {
 700             PEEKC_EOB(c, p);
 701             if (c != '\n') {
 702                 if (str)
 703                     cstr_ccat(str, '\r');
 704             } else {
 705                 file->line_num++;
 706                 goto add_char;
 707             }
 708         } else {
 709         add_char:
 710             if (str)
 711                 cstr_ccat(str, c);
 712             p++;
 713         }
 714     }
 715     p++;
 716     return p;
 717 }
 718
 719 /* skip block of text until #else, #elif or #endif. skip also pairs of
 720    #if/#endif */
 721 static void preprocess_skip(void)
 722 {
 723     int a, start_of_line, c, in_warn_or_error;
 724     uint8_t *p;
 725
 726     p = file->buf_ptr;
 727     a = 0;
 728 redo_start:
 729     start_of_line = 1;
 730     in_warn_or_error = 0;
 731     for(;;) {
 732     redo_no_start:
 733         c = *p;
 734         switch(c) {
 735         case ' ':
 736         case '\t':
 737         case '\f':
 738         case '\v':
 739         case '\r':
 740             p++;
 741             goto redo_no_start;
 742         case '\n':
 743             file->line_num++;
 744             p++;
 745             goto redo_start;
 746         case '\\':
 747             file->buf_ptr = p;
 748             c = handle_eob();
 749             if (c == CH_EOF) {
 750                 expect("#endif");
 751             } else if (c == '\\') {
 752                 ch = file->buf_ptr[0];
 753                 handle_stray_noerror();
 754             }
 755             p = file->buf_ptr;
 756             goto redo_no_start;
 757         /* skip strings */
 758         case '\"':
 759         case '\'':
 760             if (in_warn_or_error)
 761                 goto _default;
 762             p = parse_pp_string(p, c, NULL);
 763             break;
 764         /* skip comments */
 765         case '/':
 766             if (in_warn_or_error)
 767                 goto _default;
 768             file->buf_ptr = p;
 769             ch = *p;
 770             minp();
 771             p = file->buf_ptr;
 772             if (ch == '*') {
 773                 p = parse_comment(p);
 774             } else if (ch == '/') {
 775                 p = parse_line_comment(p);
 776             }
 777             break;
 778         case '#':
 779             p++;
 780             if (start_of_line) {
 781                 file->buf_ptr = p;
 782                 next_nomacro();
 783                 p = file->buf_ptr;
 784                 if (a == 0 &&
 785                     (tok == TOK_ELSE || tok == TOK_ELIF || tok == TOK_ENDIF))
 786                     goto the_end;
 787                 if (tok == TOK_IF || tok == TOK_IFDEF || tok == TOK_IFNDEF)
 788                     a++;
 789                 else if (tok == TOK_ENDIF)
 790                     a--;
 791                 else if( tok == TOK_ERROR || tok == TOK_WARNING)
 792                     in_warn_or_error = 1;
 793                 else if (tok == TOK_LINEFEED)
 794                     goto redo_start;
 795             }
 796             break;
 797 _default:
 798         default:
 799             p++;
 800             break;
 801         }
 802         start_of_line = 0;
 803     }
 804  the_end: ;
 805     file->buf_ptr = p;
 806 }
 807
 808 /* ParseState handling */
 809
 810 /* XXX: currently, no include file info is stored. Thus, we cannot display
 811    accurate messages if the function or data definition spans multiple
 812    files */
 813
 814 /* save current parse state in 's' */
 815 ST_FUNC void save_parse_state(ParseState *s)
 816 {
 817     s->line_num = file->line_num;
 818     s->macro_ptr = macro_ptr;
 819     s->tok = tok;
 820     s->tokc = tokc;
 821 }
 822
 823 /* restore parse state from 's' */
 824 ST_FUNC void restore_parse_state(ParseState *s)
 825 {
 826     file->line_num = s->line_num;
 827     macro_ptr = s->macro_ptr;
 828     tok = s->tok;
 829     tokc = s->tokc;
 830 }
 831
 832 /* return the number of additional 'ints' necessary to store the
 833    token */
 834 static inline int tok_ext_size(int t)
 835 {
 836     switch(t) {
 837         /* 4 bytes */
 838     case TOK_CINT:
 839     case TOK_CUINT:
 840     case TOK_CCHAR:
 841     case TOK_LCHAR:
 842     case TOK_CFLOAT:
 843     case TOK_LINENUM:
 844         return 1;
 845     case TOK_STR:
 846     case TOK_LSTR:
 847     case TOK_PPNUM:
 848         tcc_error("unsupported token");
 849         return 1;
 850     case TOK_CDOUBLE:
 851     case TOK_CLLONG:
 852     case TOK_CULLONG:
 853         return 2;
 854     case TOK_CLDOUBLE:
 855         return LDOUBLE_SIZE / 4;
 856     default:
 857         return 0;
 858     }
 859 }
 860
 861 /* token string handling */
 862
 863 ST_INLN void tok_str_new(TokenString *s)
 864 {
 865     s->str = NULL;
 866     s->len = 0;
 867     s->allocated_len = 0;
 868     s->last_line_num = -1;
 869 }
 870
 871 ST_FUNC void tok_str_free(int *str)
 872 {
 873     tcc_free(str);
 874 }
 875
 876 static int *tok_str_realloc(TokenString *s)
 877 {
 878     int *str, len;
 879
 880     if (s->allocated_len == 0) {
 881         len = 8;
 882     } else {
 883         len = s->allocated_len * 2;
 884     }
 885     str = tcc_realloc(s->str, len * sizeof(int));
 886     s->allocated_len = len;
 887     s->str = str;
 888     return str;
 889 }
 890
 891 ST_FUNC void tok_str_add(TokenString *s, int t)
 892 {
 893     int len, *str;
 894
 895     len = s->len;
 896     str = s->str;
 897     if (len >= s->allocated_len)
 898         str = tok_str_realloc(s);
 899     str[len++] = t;
 900     s->len = len;
 901 }
 902
 903 static void tok_str_add2(TokenString *s, int t, CValue *cv)
 904 {
 905     int len, *str;
 906
 907     len = s->len;
 908     str = s->str;
 909
 910     /* allocate space for worst case */
 911     if (len + TOK_MAX_SIZE > s->allocated_len)
 912         str = tok_str_realloc(s);
 913     str[len++] = t;
 914     switch(t) {
 915     case TOK_CINT:
 916     case TOK_CUINT:
 917     case TOK_CCHAR:
 918     case TOK_LCHAR:
 919     case TOK_CFLOAT:
 920     case TOK_LINENUM:
 921         str[len++] = cv->tab[0];
 922         break;
 923     case TOK_PPNUM:
 924     case TOK_STR:
 925     case TOK_LSTR:
 926         {
 927             int nb_words;
 928             CString *cstr;
 929
 930             nb_words = (sizeof(CString) + cv->cstr->size + 3) >> 2;
 931             while ((len + nb_words) > s->allocated_len)
 932                 str = tok_str_realloc(s);
 933             cstr = (CString *)(str + len);
 934             cstr->data = NULL;
 935             cstr->size = cv->cstr->size;
 936             cstr->data_allocated = NULL;
 937             cstr->size_allocated = cstr->size;
 938             memcpy((char *)cstr + sizeof(CString),
 939                    cv->cstr->data, cstr->size);
 940             len += nb_words;
 941         }
 942         break;
 943     case TOK_CDOUBLE:
 944     case TOK_CLLONG:
 945     case TOK_CULLONG:
 946 #if LDOUBLE_SIZE == 8
 947     case TOK_CLDOUBLE:
 948 #endif
 949         str[len++] = cv->tab[0];
 950         str[len++] = cv->tab[1];
 951         break;
 952 #if LDOUBLE_SIZE == 12
 953     case TOK_CLDOUBLE:
 954         str[len++] = cv->tab[0];
 955         str[len++] = cv->tab[1];
 956         str[len++] = cv->tab[2];
 957 #elif LDOUBLE_SIZE == 16
 958     case TOK_CLDOUBLE:
 959         str[len++] = cv->tab[0];
 960         str[len++] = cv->tab[1];
 961         str[len++] = cv->tab[2];
 962         str[len++] = cv->tab[3];
 963 #elif LDOUBLE_SIZE != 8
 964 #error add long double size support
 965 #endif
 966         break;
 967     default:
 968         break;
 969     }
 970     s->len = len;
 971 }
 972
 973 /* add the current parse token in token string 's' */
 974 ST_FUNC void tok_str_add_tok(TokenString *s)
 975 {
 976     CValue cval;
 977
 978     /* save line number info */
 979     if (file->line_num != s->last_line_num) {
 980         s->last_line_num = file->line_num;
 981         cval.i = s->last_line_num;
 982         tok_str_add2(s, TOK_LINENUM, &cval);
 983     }
 984     tok_str_add2(s, tok, &tokc);
 985 }
 986
 987 /* get a token from an integer array and increment pointer
 988    accordingly. we code it as a macro to avoid pointer aliasing. */
 989 static inline void TOK_GET(int *t, const int **pp, CValue *cv)
 990 {
 991     const int *p = *pp;
 992     int n, *tab;
 993
 994     tab = cv->tab;
 995     switch(*t = *p++) {
 996     case TOK_CINT:
 997     case TOK_CUINT:
 998     case TOK_CCHAR:
 999     case TOK_LCHAR:
1000     case TOK_CFLOAT:
1001     case TOK_LINENUM:
1002         tab[0] = *p++;
1003         break;
1004     case TOK_STR:
1005     case TOK_LSTR:
1006     case TOK_PPNUM:
1007         cv->cstr = (CString *)p;
1008         cv->cstr->data = (char *)p + sizeof(CString);
1009         p += (sizeof(CString) + cv->cstr->size + 3) >> 2;
1010         break;
1011     case TOK_CDOUBLE:
1012     case TOK_CLLONG:
1013     case TOK_CULLONG:
1014         n = 2;
1015         goto copy;
1016     case TOK_CLDOUBLE:
1017 #if LDOUBLE_SIZE == 16
1018         n = 4;
1019 #elif LDOUBLE_SIZE == 12
1020         n = 3;
1021 #elif LDOUBLE_SIZE == 8
1022         n = 2;
1023 #else
1024 # error add long double size support
1025 #endif
1026     copy:
1027         do
1028             *tab++ = *p++;
1029         while (--n);
1030         break;
1031     default:
1032         break;
1033     }
1034     *pp = p;
1035 }
1036
1037 static int macro_is_equal(const int *a, const int *b)
1038 {
1039     char buf[STRING_MAX_SIZE + 1];
1040     CValue cv;
1041     int t;
1042     while (*a && *b) {
1043         TOK_GET(&t, &a, &cv);
1044         pstrcpy(buf, sizeof buf, get_tok_str(t, &cv));
1045         TOK_GET(&t, &b, &cv);
1046         if (strcmp(buf, get_tok_str(t, &cv)))
1047             return 0;
1048     }
1049     return !(*a || *b);
1050 }
1051
1052 static void define_print(Sym *s, int is_undef)
1053 {
1054     int c, t;
1055     CValue cval;
1056     const int *str;
1057     Sym *arg;
1058
1059     if (tcc_state->dflag == 0 || !s || !tcc_state->ppfp)
1060         return;
1061
1062     if (file) {
1063         c = file->line_num - file->line_ref - 1;
1064         if (c > 0) {
1065             while (c--)
1066                 fputs("\n", tcc_state->ppfp);
1067             file->line_ref = file->line_num;
1068         }
1069     }
1070
1071     if (is_undef) {
1072         fprintf(tcc_state->ppfp, "// #undef %s\n", get_tok_str(s->v, NULL));
1073         return;
1074     }
1075
1076     fprintf(tcc_state->ppfp, "// #define %s", get_tok_str(s->v, NULL));
1077     arg = s->next;
1078     if (arg) {
1079         char *sep = "(";
1080         while (arg) {
1081             fprintf(tcc_state->ppfp, "%s%s", sep, get_tok_str(arg->v & ~SYM_FIELD, NULL));
1082             sep = ",";
1083             arg = arg->next;
1084         }
1085         fprintf(tcc_state->ppfp, ")");
1086     }
1087
1088     str = s->d;
1089     if (str)
1090         fprintf(tcc_state->ppfp, " ");
1091
1092     while (str) {
1093         TOK_GET(&t, &str, &cval);
1094         if (!t)
1095             break;
1096         fprintf(tcc_state->ppfp, "%s", get_tok_str(t, &cval));
1097     }
1098     fprintf(tcc_state->ppfp, "\n");
1099 }
1100
1101 /* defines handling */
1102 ST_INLN void define_push(int v, int macro_type, int *str, Sym *first_arg)
1103 {
1104     Sym *s;
1105
1106     s = define_find(v);
1107     if (s && !macro_is_equal(s->d, str))
1108         tcc_warning("%s redefined", get_tok_str(v, NULL));
1109
1110     s = sym_push2(&define_stack, v, macro_type, 0);
1111     s->d = str;
1112     s->next = first_arg;
1113     table_ident[v - TOK_IDENT]->sym_define = s;
1114     define_print(s, 0);
1115 }
1116
1117 /* undefined a define symbol. Its name is just set to zero */
1118 ST_FUNC void define_undef(Sym *s)
1119 {
1120     int v;
1121     v = s->v;
1122     if (v >= TOK_IDENT && v < tok_ident) {
1123         define_print(s, 1);
1124         table_ident[v - TOK_IDENT]->sym_define = NULL;
1125     }
1126     s->v = 0;
1127 }
1128
1129 ST_INLN Sym *define_find(int v)
1130 {
1131     v -= TOK_IDENT;
1132     if ((unsigned)v >= (unsigned)(tok_ident - TOK_IDENT))
1133         return NULL;
1134     return table_ident[v]->sym_define;
1135 }
1136
1137 /* free define stack until top reaches 'b' */
1138 ST_FUNC void free_defines(Sym *b)
1139 {
1140     Sym *top, *top1;
1141     int v;
1142
1143     top = define_stack;
1144     while (top != b) {
1145         top1 = top->prev;
1146         /* do not free args or predefined defines */
1147         if (top->d)
1148             tok_str_free(top->d);
1149         v = top->v;
1150         if (v >= TOK_IDENT && v < tok_ident)
1151             table_ident[v - TOK_IDENT]->sym_define = NULL;
1152         sym_free(top);
1153         top = top1;
1154     }
1155     define_stack = b;
1156 }
1157
1158 void print_defines(void)
1159 {
1160     Sym *top, *s;
1161     int v;
1162
1163     top = define_stack;
1164     while (top) {
1165         v = top->v;
1166         if (v >= TOK_IDENT && v < tok_ident) {
1167             s = table_ident[v - TOK_IDENT]->sym_define;
1168             define_print(s, 0);
1169         }
1170         top = top->prev;
1171     }
1172 }
1173
1174 /* label lookup */
1175 ST_FUNC Sym *label_find(int v)
1176 {
1177     v -= TOK_IDENT;
1178     if ((unsigned)v >= (unsigned)(tok_ident - TOK_IDENT))
1179         return NULL;
1180     return table_ident[v]->sym_label;
1181 }
1182
1183 ST_FUNC Sym *label_push(Sym **ptop, int v, int flags)
1184 {
1185     Sym *s, **ps;
1186     s = sym_push2(ptop, v, 0, 0);
1187     s->r = flags;
1188     ps = &table_ident[v - TOK_IDENT]->sym_label;
1189     if (ptop == &global_label_stack) {
1190         /* modify the top most local identifier, so that
1191            sym_identifier will point to 's' when popped */
1192         while (*ps != NULL)
1193             ps = &(*ps)->prev_tok;
1194     }
1195     s->prev_tok = *ps;
1196     *ps = s;
1197     return s;
1198 }
1199
1200 /* pop labels until element last is reached. Look if any labels are
1201    undefined. Define symbols if '&&label' was used. */
1202 ST_FUNC void label_pop(Sym **ptop, Sym *slast)
1203 {
1204     Sym *s, *s1;
1205     for(s = *ptop; s != slast; s = s1) {
1206         s1 = s->prev;
1207         if (s->r == LABEL_DECLARED) {
1208             tcc_warning("label '%s' declared but not used", get_tok_str(s->v, NULL));
1209         } else if (s->r == LABEL_FORWARD) {
1210                 tcc_error("label '%s' used but not defined",
1211                       get_tok_str(s->v, NULL));
1212         } else {
1213             if (s->c) {
1214                 /* define corresponding symbol. A size of
1215                    1 is put. */
1216                 put_extern_sym(s, cur_text_section, s->jnext, 1);
1217             }
1218         }
1219         /* remove label */
1220         table_ident[s->v - TOK_IDENT]->sym_label = s->prev_tok;
1221         sym_free(s);
1222     }
1223     *ptop = slast;
1224 }
1225
1226 /* eval an expression for #if/#elif */
1227 static int expr_preprocess(void)
1228 {
1229     int c, t;
1230     TokenString str;
1231
1232     tok_str_new(&str);
1233     while (tok != TOK_LINEFEED && tok != TOK_EOF) {
1234         next(); /* do macro subst */
1235         if (tok == TOK_DEFINED) {
1236             next_nomacro();
1237             t = tok;
1238             if (t == '(')
1239                 next_nomacro();
1240             c = define_find(tok) != 0;
1241             if (t == '(')
1242                 next_nomacro();
1243             tok = TOK_CINT;
1244             tokc.i = c;
1245         } else if (tok >= TOK_IDENT) {
1246             /* if undefined macro */
1247             tok = TOK_CINT;
1248             tokc.i = 0;
1249         }
1250         tok_str_add_tok(&str);
1251     }
1252     tok_str_add(&str, -1); /* simulate end of file */
1253     tok_str_add(&str, 0);
1254     /* now evaluate C constant expression */
1255     macro_ptr = str.str;
1256     next();
1257     c = expr_const();
1258     macro_ptr = NULL;
1259     tok_str_free(str.str);
1260     return c != 0;
1261 }
1262
1263 /* parse after #define */
1264 ST_FUNC void parse_define(void)
1265 {
1266     Sym *s, *first, **ps;
1267     int v, t, varg, is_vaargs, spc, ptok, macro_list_start;
1268     TokenString str;
1269
1270     v = tok;
1271     if (v < TOK_IDENT)
1272         tcc_error("invalid macro name '%s'", get_tok_str(tok, &tokc));
1273     /* XXX: should check if same macro (ANSI) */
1274     first = NULL;
1275     t = MACRO_OBJ;
1276     /* '(' must be just after macro definition for MACRO_FUNC */
1277     next_nomacro_spc();
1278     if (tok == '(') {
1279         next_nomacro();
1280         ps = &first;
1281         while (tok != ')') {
1282             varg = tok;
1283             next_nomacro();
1284             is_vaargs = 0;
1285             if (varg == TOK_DOTS) {
1286                 varg = TOK___VA_ARGS__;
1287                 is_vaargs = 1;
1288             } else if (tok == TOK_DOTS && gnu_ext) {
1289                 is_vaargs = 1;
1290                 next_nomacro();
1291             }
1292             if (varg < TOK_IDENT)
1293                 tcc_error( "\'%s\' may not appear in parameter list", get_tok_str(varg, NULL));
1294             s = sym_push2(&define_stack, varg | SYM_FIELD, is_vaargs, 0);
1295             *ps = s;
1296             ps = &s->next;
1297             if (tok != ',')
1298                 continue;
1299             next_nomacro();
1300         }
1301         next_nomacro_spc();
1302         t = MACRO_FUNC;
1303     }
1304     tok_str_new(&str);
1305     spc = 2;
1306     /* EOF testing necessary for '-D' handling */
1307     ptok = 0;
1308     macro_list_start = 1;
1309     while (tok != TOK_LINEFEED && tok != TOK_EOF) {
1310         if (!macro_list_start && spc == 2 && tok == TOK_TWOSHARPS)
1311             tcc_error("'##' invalid at start of macro");
1312         ptok = tok;
1313         /* remove spaces around ## and after '#' */
1314         if (TOK_TWOSHARPS == tok) {
1315             if (1 == spc)
1316                 --str.len;
1317             spc = 2;
1318         } else if ('#' == tok) {
1319             spc = 2;
1320         } else if (check_space(tok, &spc)) {
1321             goto skip;
1322         }
1323         tok_str_add2(&str, tok, &tokc);
1324     skip:
1325         next_nomacro_spc();
1326         macro_list_start = 0;
1327     }
1328     if (ptok == TOK_TWOSHARPS)
1329         tcc_error("'##' invalid at end of macro");
1330     if (spc == 1)
1331         --str.len; /* remove trailing space */
1332     tok_str_add(&str, 0);
1333     define_push(v, t, str.str, first);
1334 }
1335
1336 static inline int hash_cached_include(const char *filename)
1337 {
1338     const unsigned char *s;
1339     unsigned int h;
1340
1341     h = TOK_HASH_INIT;
1342     s = (unsigned char *) filename;
1343     while (*s) {
1344         h = TOK_HASH_FUNC(h, *s);
1345         s++;
1346     }
1347     h &= (CACHED_INCLUDES_HASH_SIZE - 1);
1348     return h;
1349 }
1350
1351 static CachedInclude *search_cached_include(TCCState *s1, const char *filename)
1352 {
1353     CachedInclude *e;
1354     int i, h;
1355     h = hash_cached_include(filename);
1356     i = s1->cached_includes_hash[h];
1357     for(;;) {
1358         if (i == 0)
1359             break;
1360         e = s1->cached_includes[i - 1];
1361         if (0 == PATHCMP(e->filename, filename))
1362             return e;
1363         i = e->hash_next;
1364     }
1365     return NULL;
1366 }
1367
1368 static inline void add_cached_include(TCCState *s1, const char *filename, int ifndef_macro)
1369 {
1370     CachedInclude *e;
1371     int h;
1372
1373     if (search_cached_include(s1, filename))
1374         return;
1375 #ifdef INC_DEBUG
1376     printf("adding cached '%s' %s\n", filename, get_tok_str(ifndef_macro, NULL));
1377 #endif
1378     e = tcc_malloc(sizeof(CachedInclude) + strlen(filename));
1379     strcpy(e->filename, filename);
1380     e->ifndef_macro = ifndef_macro;
1381     dynarray_add((void ***)&s1->cached_includes, &s1->nb_cached_includes, e);
1382     /* add in hash table */
1383     h = hash_cached_include(filename);
1384     e->hash_next = s1->cached_includes_hash[h];
1385     s1->cached_includes_hash[h] = s1->nb_cached_includes;
1386 }
1387
1388 static void pragma_parse(TCCState *s1)
1389 {
1390     int val;
1391
1392     next();
1393     if (tok == TOK_pack) {
1394         /*
1395           This may be:
1396           #pragma pack(1) // set
1397           #pragma pack() // reset to default
1398           #pragma pack(push,1) // push & set
1399           #pragma pack(pop) // restore previous
1400         */
1401         next();
1402         skip('(');
1403         if (tok == TOK_ASM_pop) {
1404             next();
1405             if (s1->pack_stack_ptr <= s1->pack_stack) {
1406             stk_error:
1407                 tcc_error("out of pack stack");
1408             }
1409             s1->pack_stack_ptr--;
1410         } else {
1411             val = 0;
1412             if (tok != ')') {
1413                 if (tok == TOK_ASM_push) {
1414                     next();
1415                     if (s1->pack_stack_ptr >= s1->pack_stack + PACK_STACK_SIZE - 1)
1416                         goto stk_error;
1417                     s1->pack_stack_ptr++;
1418                     skip(',');
1419                 }
1420                 if (tok != TOK_CINT) {
1421                 pack_error:
1422                     tcc_error("invalid pack pragma");
1423                 }
1424                 val = tokc.i;
1425                 if (val < 1 || val > 16 || (val & (val - 1)) != 0)
1426                     goto pack_error;
1427                 next();
1428             }
1429             *s1->pack_stack_ptr = val;
1430             skip(')');
1431         }
1432     }
1433 }
1434
1435 /* is_bof is true if first non space token at beginning of file */
1436 ST_FUNC void preprocess(int is_bof)
1437 {
1438     TCCState *s1 = tcc_state;
1439     int i, c, n, saved_parse_flags;
1440     char buf[1024], *q;
1441     Sym *s;
1442
1443     saved_parse_flags = parse_flags;
1444     parse_flags = PARSE_FLAG_PREPROCESS | PARSE_FLAG_TOK_NUM |
1445         PARSE_FLAG_LINEFEED;
1446     next_nomacro();
1447  redo:
1448     switch(tok) {
1449     case TOK_DEFINE:
1450         next_nomacro();
1451         parse_define();
1452         break;
1453     case TOK_UNDEF:
1454         next_nomacro();
1455         s = define_find(tok);
1456         /* undefine symbol by putting an invalid name */
1457         if (s)
1458             define_undef(s);
1459         break;
1460     case TOK_INCLUDE:
1461     case TOK_INCLUDE_NEXT:
1462         ch = file->buf_ptr[0];
1463         /* XXX: incorrect if comments : use next_nomacro with a special mode */
1464         skip_spaces();
1465         if (ch == '<') {
1466             c = '>';
1467             goto read_name;
1468         } else if (ch == '\"') {
1469             c = ch;
1470         read_name:
1471             inp();
1472             q = buf;
1473             while (ch != c && ch != '\n' && ch != CH_EOF) {
1474                 if ((q - buf) < sizeof(buf) - 1)
1475                     *q++ = ch;
1476                 if (ch == '\\') {
1477                     if (handle_stray_noerror() == 0)
1478                         --q;
1479                 } else
1480                     inp();
1481             }
1482             *q = '\0';
1483             minp();
1484 #if 0
1485             /* eat all spaces and comments after include */
1486             /* XXX: slightly incorrect */
1487             while (ch1 != '\n' && ch1 != CH_EOF)
1488                 inp();
1489 #endif
1490         } else {
1491             /* computed #include : either we have only strings or
1492                we have anything enclosed in '<>' */
1493             next();
1494             buf[0] = '\0';
1495             if (tok == TOK_STR) {
1496                 while (tok != TOK_LINEFEED) {
1497                     if (tok != TOK_STR) {
1498                     include_syntax:
1499                         tcc_error("'#include' expects \"FILENAME\" or <FILENAME>");
1500                     }
1501                     pstrcat(buf, sizeof(buf), (char *)tokc.cstr->data);
1502                     next();
1503                 }
1504                 c = '\"';
1505             } else {
1506                 int len;
1507                 while (tok != TOK_LINEFEED) {
1508                     pstrcat(buf, sizeof(buf), get_tok_str(tok, &tokc));
1509                     next();
1510                 }
1511                 len = strlen(buf);
1512                 /* check syntax and remove '<>' */
1513                 if (len < 2 || buf[0] != '<' || buf[len - 1] != '>')
1514                     goto include_syntax;
1515                 memmove(buf, buf + 1, len - 2);
1516                 buf[len - 2] = '\0';
1517                 c = '>';
1518             }
1519         }
1520
1521         if (s1->include_stack_ptr >= s1->include_stack + INCLUDE_STACK_SIZE)
1522             tcc_error("#include recursion too deep");
1523         /* store current file in stack, but increment stack later below */
1524         *s1->include_stack_ptr = file;
1525
1526         n = s1->nb_include_paths + s1->nb_sysinclude_paths;
1527         for (i = -2; i < n; ++i) {
1528             char buf1[sizeof file->filename];
1529             CachedInclude *e;
1530             BufferedFile **f;
1531             const char *path;
1532
1533             if (i == -2) {
1534                 /* check absolute include path */
1535                 if (!IS_ABSPATH(buf))
1536                     continue;
1537                 buf1[0] = 0;
1538                 i = n; /* force end loop */
1539
1540             } else if (i == -1) {
1541                 /* search in current dir if "header.h" */
1542                 if (c != '\"')
1543                     continue;
1544                 path = file->filename;
1545                 pstrncpy(buf1, path, tcc_basename(path) - path);
1546
1547             } else {
1548                 /* search in all the include paths */
1549                 if (i < s1->nb_include_paths)
1550                     path = s1->include_paths[i];
1551                 else
1552                     path = s1->sysinclude_paths[i - s1->nb_include_paths];
1553                 pstrcpy(buf1, sizeof(buf1), path);
1554                 pstrcat(buf1, sizeof(buf1), "/");
1555             }
1556
1557             pstrcat(buf1, sizeof(buf1), buf);
1558
1559             if (tok == TOK_INCLUDE_NEXT)
1560                 for (f = s1->include_stack_ptr; f >= s1->include_stack; --f)
1561                     if (0 == PATHCMP((*f)->filename, buf1)) {
1562 #ifdef INC_DEBUG
1563                         printf("%s: #include_next skipping %s\n", file->filename, buf1);
1564 #endif
1565                         goto include_trynext;
1566                     }
1567
1568             e = search_cached_include(s1, buf1);
1569             if (e && define_find(e->ifndef_macro)) {
1570                 /* no need to parse the include because the 'ifndef macro'
1571                    is defined */
1572 #ifdef INC_DEBUG
1573                 printf("%s: skipping cached %s\n", file->filename, buf1);
1574 #endif
1575                 goto include_done;
1576             }
1577
1578             if (tcc_open(s1, buf1) < 0)
1579 include_trynext:
1580                 continue;
1581
1582 #ifdef INC_DEBUG
1583             printf("%s: including %s\n", file->prev->filename, file->filename);
1584 #endif
1585             /* update target deps */
1586             dynarray_add((void ***)&s1->target_deps, &s1->nb_target_deps,
1587                     tcc_strdup(buf1));
1588             /* push current file in stack */
1589             ++s1->include_stack_ptr;
1590             /* add include file debug info */
1591             if (s1->do_debug)
1592                 put_stabs(file->filename, N_BINCL, 0, 0, 0);
1593             tok_flags |= TOK_FLAG_BOF | TOK_FLAG_BOL;
1594             ch = file->buf_ptr[0];
1595             goto the_end;
1596         }
1597         tcc_error("include file '%s' not found", buf);
1598 include_done:
1599         break;
1600     case TOK_IFNDEF:
1601         c = 1;
1602         goto do_ifdef;
1603     case TOK_IF:
1604         c = expr_preprocess();
1605         goto do_if;
1606     case TOK_IFDEF:
1607         c = 0;
1608     do_ifdef:
1609         next_nomacro();
1610         if (tok < TOK_IDENT)
1611             tcc_error("invalid argument for '#if%sdef'", c ? "n" : "");
1612         if (is_bof) {
1613             if (c) {
1614 #ifdef INC_DEBUG
1615                 printf("#ifndef %s\n", get_tok_str(tok, NULL));
1616 #endif
1617                 file->ifndef_macro = tok;
1618             }
1619         }
1620         c = (define_find(tok) != 0) ^ c;
1621     do_if:
1622         if (s1->ifdef_stack_ptr >= s1->ifdef_stack + IFDEF_STACK_SIZE)
1623             tcc_error("memory full (ifdef)");
1624         *s1->ifdef_stack_ptr++ = c;
1625         goto test_skip;
1626     case TOK_ELSE:
1627         if (s1->ifdef_stack_ptr == s1->ifdef_stack)
1628             tcc_error("#else without matching #if");
1629         if (s1->ifdef_stack_ptr[-1] & 2)
1630             tcc_error("#else after #else");
1631         c = (s1->ifdef_stack_ptr[-1] ^= 3);
1632         goto test_else;
1633     case TOK_ELIF:
1634         if (s1->ifdef_stack_ptr == s1->ifdef_stack)
1635             tcc_error("#elif without matching #if");
1636         c = s1->ifdef_stack_ptr[-1];
1637         if (c > 1)
1638             tcc_error("#elif after #else");
1639         /* last #if/#elif expression was true: we skip */
1640         if (c == 1)
1641             goto skip;
1642         c = expr_preprocess();
1643         s1->ifdef_stack_ptr[-1] = c;
1644     test_else:
1645         if (s1->ifdef_stack_ptr == file->ifdef_stack_ptr + 1)
1646             file->ifndef_macro = 0;
1647     test_skip:
1648         if (!(c & 1)) {
1649         skip:
1650             preprocess_skip();
1651             is_bof = 0;
1652             goto redo;
1653         }
1654         break;
1655     case TOK_ENDIF:
1656         if (s1->ifdef_stack_ptr <= file->ifdef_stack_ptr)
1657             tcc_error("#endif without matching #if");
1658         s1->ifdef_stack_ptr--;
1659         /* '#ifndef macro' was at the start of file. Now we check if
1660            an '#endif' is exactly at the end of file */
1661         if (file->ifndef_macro &&
1662             s1->ifdef_stack_ptr == file->ifdef_stack_ptr) {
1663             file->ifndef_macro_saved = file->ifndef_macro;
1664             /* need to set to zero to avoid false matches if another
1665                #ifndef at middle of file */
1666             file->ifndef_macro = 0;
1667             while (tok != TOK_LINEFEED)
1668                 next_nomacro();
1669             tok_flags |= TOK_FLAG_ENDIF;
1670             goto the_end;
1671         }
1672         break;
1673     case TOK_LINE:
1674         next();
1675         if (tok != TOK_CINT)
1676             tcc_error("A #line format is wrong");
1677     case TOK_PPNUM:
1678         if (tok != TOK_CINT) {
1679             char *p = tokc.cstr->data;
1680             tokc.i = strtoul(p, (char **)&p, 10);
1681         }
1682         file->line_num = tokc.i - 1; /* the line number will be incremented after */
1683         next();
1684         if (tok != TOK_LINEFEED) {
1685             if (tok != TOK_STR)
1686                 tcc_error("#line");
1687             pstrcpy(file->filename, sizeof(file->filename),
1688                     (char *)tokc.cstr->data);
1689         }
1690         break;
1691     case TOK_ERROR:
1692     case TOK_WARNING:
1693         c = tok;
1694         ch = file->buf_ptr[0];
1695         skip_spaces();
1696         q = buf;
1697         while (ch != '\n' && ch != CH_EOF) {
1698             if ((q - buf) < sizeof(buf) - 1)
1699                 *q++ = ch;
1700             if (ch == '\\') {
1701                 if (handle_stray_noerror() == 0)
1702                     --q;
1703             } else
1704                 inp();
1705         }
1706         *q = '\0';
1707         if (c == TOK_ERROR)
1708             tcc_error("#error %s", buf);
1709         else
1710             tcc_warning("#warning %s", buf);
1711         break;
1712     case TOK_PRAGMA:
1713         pragma_parse(s1);
1714         break;
1715     default:
1716         if (tok == TOK_LINEFEED || tok == '!' || tok == TOK_PPNUM) {
1717             /* '!' is ignored to allow C scripts. numbers are ignored
1718                to emulate cpp behaviour */
1719         } else {
1720             if (!(saved_parse_flags & PARSE_FLAG_ASM_COMMENTS))
1721                 tcc_warning("Ignoring unknown preprocessing directive #%s", get_tok_str(tok, &tokc));
1722             else {
1723                 /* this is a gas line comment in an 'S' file. */
1724                 file->buf_ptr = parse_line_comment(file->buf_ptr);
1725                 goto the_end;
1726             }
1727         }
1728         break;
1729     }
1730     /* ignore other preprocess commands or #! for C scripts */
1731     while (tok != TOK_LINEFEED)
1732         next_nomacro();
1733  the_end:
1734     parse_flags = saved_parse_flags;
1735 }
1736
1737 /* evaluate escape codes in a string. */
1738 static void parse_escape_string(CString *outstr, const uint8_t *buf, int is_long)
1739 {
1740     int c, n;
1741     const uint8_t *p;
1742
1743     p = buf;
1744     for(;;) {
1745         c = *p;
1746         if (c == '\0')
1747             break;
1748         if (c == '\\') {
1749             p++;
1750             /* escape */
1751             c = *p;
1752             switch(c) {
1753             case '0': case '1': case '2': case '3':
1754             case '4': case '5': case '6': case '7':
1755                 /* at most three octal digits */
1756                 n = c - '0';
1757                 p++;
1758                 c = *p;
1759                 if (isoct(c)) {
1760                     n = n * 8 + c - '0';
1761                     p++;
1762                     c = *p;
1763                     if (isoct(c)) {
1764                         n = n * 8 + c - '0';
1765                         p++;
1766                     }
1767                 }
1768                 c = n;
1769                 goto add_char_nonext;
1770             case 'x':
1771             case 'u':
1772             case 'U':
1773                 p++;
1774                 n = 0;
1775                 for(;;) {
1776                     c = *p;
1777                     if (c >= 'a' && c <= 'f')
1778                         c = c - 'a' + 10;
1779                     else if (c >= 'A' && c <= 'F')
1780                         c = c - 'A' + 10;
1781                     else if (isnum(c))
1782                         c = c - '0';
1783                     else
1784                         break;
1785                     n = n * 16 + c;
1786                     p++;
1787                 }
1788                 c = n;
1789                 goto add_char_nonext;
1790             case 'a':
1791                 c = '\a';
1792                 break;
1793             case 'b':
1794                 c = '\b';
1795                 break;
1796             case 'f':
1797                 c = '\f';
1798                 break;
1799             case 'n':
1800                 c = '\n';
1801                 break;
1802             case 'r':
1803                 c = '\r';
1804                 break;
1805             case 't':
1806                 c = '\t';
1807                 break;
1808             case 'v':
1809                 c = '\v';
1810                 break;
1811             case 'e':
1812                 if (!gnu_ext)
1813                     goto invalid_escape;
1814                 c = 27;
1815                 break;
1816             case '\'':
1817             case '\"':
1818             case '\\':
1819             case '?':
1820                 break;
1821             default:
1822             invalid_escape:
1823                 if (c >= '!' && c <= '~')
1824                     tcc_warning("unknown escape sequence: \'\\%c\'", c);
1825                 else
1826                     tcc_warning("unknown escape sequence: \'\\x%x\'", c);
1827                 break;
1828             }
1829         }
1830         p++;
1831     add_char_nonext:
1832         if (!is_long)
1833             cstr_ccat(outstr, c);
1834         else
1835             cstr_wccat(outstr, c);
1836     }
1837     /* add a trailing '\0' */
1838     if (!is_long)
1839         cstr_ccat(outstr, '\0');
1840     else
1841         cstr_wccat(outstr, '\0');
1842 }
1843
1844 /* we use 64 bit numbers */
1845 #define BN_SIZE 2
1846
1847 /* bn = (bn << shift) | or_val */
1848 static void bn_lshift(unsigned int *bn, int shift, int or_val)
1849 {
1850     int i;
1851     unsigned int v;
1852     for(i=0;i<BN_SIZE;i++) {
1853         v = bn[i];
1854         bn[i] = (v << shift) | or_val;
1855         or_val = v >> (32 - shift);
1856     }
1857 }
1858
1859 static void bn_zero(unsigned int *bn)
1860 {
1861     int i;
1862     for(i=0;i<BN_SIZE;i++) {
1863         bn[i] = 0;
1864     }
1865 }
1866
1867 /* parse number in null terminated string 'p' and return it in the
1868    current token */
1869 static void parse_number(const char *p)
1870 {
1871     int b, t, shift, frac_bits, s, exp_val, ch;
1872     char *q;
1873     unsigned int bn[BN_SIZE];
1874     double d;
1875
1876     /* number */
1877     q = token_buf;
1878     ch = *p++;
1879     t = ch;
1880     ch = *p++;
1881     *q++ = t;
1882     b = 10;
1883     if (t == '.') {
1884         goto float_frac_parse;
1885     } else if (t == '0') {
1886         if (ch == 'x' || ch == 'X') {
1887             q--;
1888             ch = *p++;
1889             b = 16;
1890         } else if (tcc_ext && (ch == 'b' || ch == 'B')) {
1891             q--;
1892             ch = *p++;
1893             b = 2;
1894         }
1895     }
1896     /* parse all digits. cannot check octal numbers at this stage
1897        because of floating point constants */
1898     while (1) {
1899         if (ch >= 'a' && ch <= 'f')
1900             t = ch - 'a' + 10;
1901         else if (ch >= 'A' && ch <= 'F')
1902             t = ch - 'A' + 10;
1903         else if (isnum(ch))
1904             t = ch - '0';
1905         else
1906             break;
1907         if (t >= b)
1908             break;
1909         if (q >= token_buf + STRING_MAX_SIZE) {
1910         num_too_long:
1911             tcc_error("number too long");
1912         }
1913         *q++ = ch;
1914         ch = *p++;
1915     }
1916     if (ch == '.' ||
1917         ((ch == 'e' || ch == 'E') && b == 10) ||
1918         ((ch == 'p' || ch == 'P') && (b == 16 || b == 2))) {
1919         if (b != 10) {
1920             /* NOTE: strtox should support that for hexa numbers, but
1921                non ISOC99 libcs do not support it, so we prefer to do
1922                it by hand */
1923             /* hexadecimal or binary floats */
1924             /* XXX: handle overflows */
1925             *q = '\0';
1926             if (b == 16)
1927                 shift = 4;
1928             else
1929                 shift = 1;
1930             bn_zero(bn);
1931             q = token_buf;
1932             while (1) {
1933                 t = *q++;
1934                 if (t == '\0') {
1935                     break;
1936                 } else if (t >= 'a') {
1937                     t = t - 'a' + 10;
1938                 } else if (t >= 'A') {
1939                     t = t - 'A' + 10;
1940                 } else {
1941                     t = t - '0';
1942                 }
1943                 bn_lshift(bn, shift, t);
1944             }
1945             frac_bits = 0;
1946             if (ch == '.') {
1947                 ch = *p++;
1948                 while (1) {
1949                     t = ch;
1950                     if (t >= 'a' && t <= 'f') {
1951                         t = t - 'a' + 10;
1952                     } else if (t >= 'A' && t <= 'F') {
1953                         t = t - 'A' + 10;
1954                     } else if (t >= '0' && t <= '9') {
1955                         t = t - '0';
1956                     } else {
1957                         break;
1958                     }
1959                     if (t >= b)
1960                         tcc_error("invalid digit");
1961                     bn_lshift(bn, shift, t);
1962                     frac_bits += shift;
1963                     ch = *p++;
1964                 }
1965             }
1966             if (ch != 'p' && ch != 'P')
1967                 expect("exponent");
1968             ch = *p++;
1969             s = 1;
1970             exp_val = 0;
1971             if (ch == '+') {
1972                 ch = *p++;
1973             } else if (ch == '-') {
1974                 s = -1;
1975                 ch = *p++;
1976             }
1977             if (ch < '0' || ch > '9')
1978                 expect("exponent digits");
1979             while (ch >= '0' && ch <= '9') {
1980                 exp_val = exp_val * 10 + ch - '0';
1981                 ch = *p++;
1982             }
1983             exp_val = exp_val * s;
1984
1985             /* now we can generate the number */
1986             /* XXX: should patch directly float number */
1987             d = (double)bn[1] * 4294967296.0 + (double)bn[0];
1988             d = ldexp(d, exp_val - frac_bits);
1989             t = toup(ch);
1990             if (t == 'F') {
1991                 ch = *p++;
1992                 tok = TOK_CFLOAT;
1993                 /* float : should handle overflow */
1994                 tokc.f = (float)d;
1995             } else if (t == 'L') {
1996                 ch = *p++;
1997 #ifdef TCC_TARGET_PE
1998                 tok = TOK_CDOUBLE;
1999                 tokc.d = d;
2000 #else
2001                 tok = TOK_CLDOUBLE;
2002                 /* XXX: not large enough */
2003                 tokc.ld = (long double)d;
2004 #endif
2005             } else {
2006                 tok = TOK_CDOUBLE;
2007                 tokc.d = d;
2008             }
2009         } else {
2010             /* decimal floats */
2011             if (ch == '.') {
2012                 if (q >= token_buf + STRING_MAX_SIZE)
2013                     goto num_too_long;
2014                 *q++ = ch;
2015                 ch = *p++;
2016             float_frac_parse:
2017                 while (ch >= '0' && ch <= '9') {
2018                     if (q >= token_buf + STRING_MAX_SIZE)
2019                         goto num_too_long;
2020                     *q++ = ch;
2021                     ch = *p++;
2022                 }
2023             }
2024             if (ch == 'e' || ch == 'E') {
2025                 if (q >= token_buf + STRING_MAX_SIZE)
2026                     goto num_too_long;
2027                 *q++ = ch;
2028                 ch = *p++;
2029                 if (ch == '-' || ch == '+') {
2030                     if (q >= token_buf + STRING_MAX_SIZE)
2031                         goto num_too_long;
2032                     *q++ = ch;
2033                     ch = *p++;
2034                 }
2035                 if (ch < '0' || ch > '9')
2036                     expect("exponent digits");
2037                 while (ch >= '0' && ch <= '9') {
2038                     if (q >= token_buf + STRING_MAX_SIZE)
2039                         goto num_too_long;
2040                     *q++ = ch;
2041                     ch = *p++;
2042                 }
2043             }
2044             *q = '\0';
2045             t = toup(ch);
2046             errno = 0;
2047             if (t == 'F') {
2048                 ch = *p++;
2049                 tok = TOK_CFLOAT;
2050                 tokc.f = strtof(token_buf, NULL);
2051             } else if (t == 'L') {
2052                 ch = *p++;
2053 #ifdef TCC_TARGET_PE
2054                 tok = TOK_CDOUBLE;
2055                 tokc.d = strtod(token_buf, NULL);
2056 #else
2057                 tok = TOK_CLDOUBLE;
2058                 tokc.ld = strtold(token_buf, NULL);
2059 #endif
2060             } else {
2061                 tok = TOK_CDOUBLE;
2062                 tokc.d = strtod(token_buf, NULL);
2063             }
2064         }
2065     } else {
2066         unsigned long long n, n1;
2067         int lcount, ucount, must_64bit;
2068         const char *p1;
2069
2070         /* integer number */
2071         *q = '\0';
2072         q = token_buf;
2073         if (b == 10 && *q == '0') {
2074             b = 8;
2075             q++;
2076         }
2077         n = 0;
2078         while(1) {
2079             t = *q++;
2080             /* no need for checks except for base 10 / 8 errors */
2081             if (t == '\0')
2082                 break;
2083             else if (t >= 'a')
2084                 t = t - 'a' + 10;
2085             else if (t >= 'A')
2086                 t = t - 'A' + 10;
2087             else
2088                 t = t - '0';
2089             if (t >= b)
2090                 tcc_error("invalid digit");
2091             n1 = n;
2092             n = n * b + t;
2093             /* detect overflow */
2094             /* XXX: this test is not reliable */
2095             if (n < n1)
2096                 tcc_error("integer constant overflow");
2097         }
2098
2099         /* Determine the characteristics (unsigned and/or 64bit) the type of
2100            the constant must have according to the constant suffix(es) */
2101         lcount = ucount = must_64bit = 0;
2102         p1 = p;
2103         for(;;) {
2104             t = toup(ch);
2105             if (t == 'L') {
2106                 if (lcount >= 2)
2107                     tcc_error("three 'l's in integer constant");
2108                 if (lcount && *(p - 1) != ch)
2109                     tcc_error("incorrect integer suffix: %s", p1);
2110                 lcount++;
2111 #if !defined TCC_TARGET_X86_64 || defined TCC_TARGET_PE
2112                 if (lcount == 2)
2113 #endif
2114                     must_64bit = 1;
2115                 ch = *p++;
2116             } else if (t == 'U') {
2117                 if (ucount >= 1)
2118                     tcc_error("two 'u's in integer constant");
2119                 ucount++;
2120                 ch = *p++;
2121             } else {
2122                 break;
2123             }
2124         }
2125
2126         /* Whether 64 bits are needed to hold the constant's value */
2127         if (n & 0xffffffff00000000LL || must_64bit) {
2128             tok = TOK_CLLONG;
2129             n1 = n >> 32;
2130         } else {
2131             tok = TOK_CINT;
2132             n1 = n;
2133         }
2134
2135         /* Whether type must be unsigned to hold the constant's value */
2136         if (ucount || ((n1 >> 31) && (b != 10))) {
2137             if (tok == TOK_CLLONG)
2138                 tok = TOK_CULLONG;
2139             else
2140                 tok = TOK_CUINT;
2141         /* If decimal and no unsigned suffix, bump to 64 bits or throw error */
2142         } else if (n1 >> 31) {
2143             if (tok == TOK_CINT)
2144                 tok = TOK_CLLONG;
2145             else
2146                 tcc_error("integer constant overflow");
2147         }
2148
2149         if (tok == TOK_CINT || tok == TOK_CUINT)
2150             tokc.ui = n;
2151         else
2152             tokc.ull = n;
2153     }
2154     if (ch)
2155         tcc_error("invalid number\n");
2156 }
2157
2158
2159 #define PARSE2(c1, tok1, c2, tok2)              \
2160     case c1:                                    \
2161         PEEKC(c, p);                            \
2162         if (c == c2) {                          \
2163             p++;                                \
2164             tok = tok2;                         \
2165         } else {                                \
2166             tok = tok1;                         \
2167         }                                       \
2168         break;
2169
2170 /* return next token without macro substitution */
2171 static inline void next_nomacro1(void)
2172 {
2173     int t, c, is_long;
2174     TokenSym *ts;
2175     uint8_t *p, *p1;
2176     unsigned int h;
2177
2178     p = file->buf_ptr;
2179  redo_no_start:
2180     c = *p;
2181     switch(c) {
2182     case ' ':
2183     case '\t':
2184         tok = c;
2185         p++;
2186         goto keep_tok_flags;
2187     case '\f':
2188     case '\v':
2189     case '\r':
2190         p++;
2191         goto redo_no_start;
2192     case '\\':
2193         /* first look if it is in fact an end of buffer */
2194         if (p >= file->buf_end) {
2195             file->buf_ptr = p;
2196             handle_eob();
2197             p = file->buf_ptr;
2198             if (p >= file->buf_end)
2199                 goto parse_eof;
2200             else
2201                 goto redo_no_start;
2202         } else {
2203             file->buf_ptr = p;
2204             ch = *p;
2205             handle_stray();
2206             p = file->buf_ptr;
2207             goto redo_no_start;
2208         }
2209     parse_eof:
2210         {
2211             TCCState *s1 = tcc_state;
2212             if ((parse_flags & PARSE_FLAG_LINEFEED)
2213                 && !(tok_flags & TOK_FLAG_EOF)) {
2214                 tok_flags |= TOK_FLAG_EOF;
2215                 tok = TOK_LINEFEED;
2216                 goto keep_tok_flags;
2217             } else if (!(parse_flags & PARSE_FLAG_PREPROCESS)) {
2218                 tok = TOK_EOF;
2219             } else if (s1->ifdef_stack_ptr != file->ifdef_stack_ptr) {
2220                 tcc_error("missing #endif");
2221             } else if (s1->include_stack_ptr == s1->include_stack) {
2222                 /* no include left : end of file. */
2223                 tok = TOK_EOF;
2224             } else {
2225                 tok_flags &= ~TOK_FLAG_EOF;
2226                 /* pop include file */
2227
2228                 /* test if previous '#endif' was after a #ifdef at
2229                    start of file */
2230                 if (tok_flags & TOK_FLAG_ENDIF) {
2231 #ifdef INC_DEBUG
2232                     printf("#endif %s\n", get_tok_str(file->ifndef_macro_saved, NULL));
2233 #endif
2234                     add_cached_include(s1, file->filename, file->ifndef_macro_saved);
2235                     tok_flags &= ~TOK_FLAG_ENDIF;
2236                 }
2237
2238                 /* add end of include file debug info */
2239                 if (tcc_state->do_debug) {
2240                     put_stabd(N_EINCL, 0, 0);
2241                 }
2242                 /* pop include stack */
2243                 tcc_close();
2244                 s1->include_stack_ptr--;
2245                 p = file->buf_ptr;
2246                 goto redo_no_start;
2247             }
2248         }
2249         break;
2250
2251     case '\n':
2252         file->line_num++;
2253         tok_flags |= TOK_FLAG_BOL;
2254         p++;
2255 maybe_newline:
2256         if (0 == (parse_flags & PARSE_FLAG_LINEFEED))
2257             goto redo_no_start;
2258         tok = TOK_LINEFEED;
2259         goto keep_tok_flags;
2260
2261     case '#':
2262         /* XXX: simplify */
2263         PEEKC(c, p);
2264         if ((tok_flags & TOK_FLAG_BOL) &&
2265             (parse_flags & PARSE_FLAG_PREPROCESS)) {
2266             file->buf_ptr = p;
2267             preprocess(tok_flags & TOK_FLAG_BOF);
2268             p = file->buf_ptr;
2269             goto maybe_newline;
2270         } else {
2271             if (c == '#') {
2272                 p++;
2273                 tok = TOK_TWOSHARPS;
2274             } else {
2275                 if (parse_flags & PARSE_FLAG_ASM_COMMENTS) {
2276                     p = parse_line_comment(p - 1);
2277                     goto redo_no_start;
2278                 } else {
2279                     tok = '#';
2280                 }
2281             }
2282         }
2283         break;
2284
2285     case 'a': case 'b': case 'c': case 'd':
2286     case 'e': case 'f': case 'g': case 'h':
2287     case 'i': case 'j': case 'k': case 'l':
2288     case 'm': case 'n': case 'o': case 'p':
2289     case 'q': case 'r': case 's': case 't':
2290     case 'u': case 'v': case 'w': case 'x':
2291     case 'y': case 'z':
2292     case 'A': case 'B': case 'C': case 'D':
2293     case 'E': case 'F': case 'G': case 'H':
2294     case 'I': case 'J': case 'K':
2295     case 'M': case 'N': case 'O': case 'P':
2296     case 'Q': case 'R': case 'S': case 'T':
2297     case 'U': case 'V': case 'W': case 'X':
2298     case 'Y': case 'Z':
2299     case '_':
2300     parse_ident_fast:
2301         p1 = p;
2302         h = TOK_HASH_INIT;
2303         h = TOK_HASH_FUNC(h, c);
2304         p++;
2305         for(;;) {
2306             c = *p;
2307             if (!isidnum_table[c-CH_EOF])
2308                 break;
2309             h = TOK_HASH_FUNC(h, c);
2310             p++;
2311         }
2312         if (c != '\\') {
2313             TokenSym **pts;
2314             int len;
2315
2316             /* fast case : no stray found, so we have the full token
2317                and we have already hashed it */
2318             len = p - p1;
2319             h &= (TOK_HASH_SIZE - 1);
2320             pts = &hash_ident[h];
2321             for(;;) {
2322                 ts = *pts;
2323                 if (!ts)
2324                     break;
2325                 if (ts->len == len && !memcmp(ts->str, p1, len))
2326                     goto token_found;
2327                 pts = &(ts->hash_next);
2328             }
2329             ts = tok_alloc_new(pts, (char *) p1, len);
2330         token_found: ;
2331         } else {
2332             /* slower case */
2333             cstr_reset(&tokcstr);
2334
2335             while (p1 < p) {
2336                 cstr_ccat(&tokcstr, *p1);
2337                 p1++;
2338             }
2339             p--;
2340             PEEKC(c, p);
2341         parse_ident_slow:
2342             while (isidnum_table[c-CH_EOF]) {
2343                 cstr_ccat(&tokcstr, c);
2344                 PEEKC(c, p);
2345             }
2346             ts = tok_alloc(tokcstr.data, tokcstr.size);
2347         }
2348         tok = ts->tok;
2349         break;
2350     case 'L':
2351         t = p[1];
2352         if (t != '\\' && t != '\'' && t != '\"') {
2353             /* fast case */
2354             goto parse_ident_fast;
2355         } else {
2356             PEEKC(c, p);
2357             if (c == '\'' || c == '\"') {
2358                 is_long = 1;
2359                 goto str_const;
2360             } else {
2361                 cstr_reset(&tokcstr);
2362                 cstr_ccat(&tokcstr, 'L');
2363                 goto parse_ident_slow;
2364             }
2365         }
2366         break;
2367     case '0': case '1': case '2': case '3':
2368     case '4': case '5': case '6': case '7':
2369     case '8': case '9':
2370
2371         cstr_reset(&tokcstr);
2372         /* after the first digit, accept digits, alpha, '.' or sign if
2373            prefixed by 'eEpP' */
2374     parse_num:
2375         for(;;) {
2376             t = c;
2377             cstr_ccat(&tokcstr, c);
2378             PEEKC(c, p);
2379             if (!(isnum(c) || isid(c) || c == '.' ||
2380                   ((c == '+' || c == '-') &&
2381                    (t == 'e' || t == 'E' || t == 'p' || t == 'P'))))
2382                 break;
2383         }
2384         /* We add a trailing '\0' to ease parsing */
2385         cstr_ccat(&tokcstr, '\0');
2386         tokc.cstr = &tokcstr;
2387         tok = TOK_PPNUM;
2388         break;
2389     case '.':
2390         /* special dot handling because it can also start a number */
2391         PEEKC(c, p);
2392         if (isnum(c)) {
2393             cstr_reset(&tokcstr);
2394             cstr_ccat(&tokcstr, '.');
2395             goto parse_num;
2396         } else if (c == '.') {
2397             PEEKC(c, p);
2398             if (c != '.')
2399                 expect("'.'");
2400             PEEKC(c, p);
2401             tok = TOK_DOTS;
2402         } else {
2403             tok = '.';
2404         }
2405         break;
2406     case '\'':
2407     case '\"':
2408         is_long = 0;
2409     str_const:
2410         {
2411             CString str;
2412             int sep;
2413
2414             sep = c;
2415
2416             /* parse the string */
2417             cstr_new(&str);
2418             p = parse_pp_string(p, sep, &str);
2419             cstr_ccat(&str, '\0');
2420
2421             /* eval the escape (should be done as TOK_PPNUM) */
2422             cstr_reset(&tokcstr);
2423             parse_escape_string(&tokcstr, str.data, is_long);
2424             cstr_free(&str);
2425
2426             if (sep == '\'') {
2427                 int char_size;
2428                 /* XXX: make it portable */
2429                 if (!is_long)
2430                     char_size = 1;
2431                 else
2432                     char_size = sizeof(nwchar_t);
2433                 if (tokcstr.size <= char_size)
2434                     tcc_error("empty character constant");
2435                 if (tokcstr.size > 2 * char_size)
2436                     tcc_warning("multi-character character constant");
2437                 if (!is_long) {
2438                     tokc.i = *(int8_t *)tokcstr.data;
2439                     tok = TOK_CCHAR;
2440                 } else {
2441                     tokc.i = *(nwchar_t *)tokcstr.data;
2442                     tok = TOK_LCHAR;
2443                 }
2444             } else {
2445                 tokc.cstr = &tokcstr;
2446                 if (!is_long)
2447                     tok = TOK_STR;
2448                 else
2449                     tok = TOK_LSTR;
2450             }
2451         }
2452         break;
2453
2454     case '<':
2455         PEEKC(c, p);
2456         if (c == '=') {
2457             p++;
2458             tok = TOK_LE;
2459         } else if (c == '<') {
2460             PEEKC(c, p);
2461             if (c == '=') {
2462                 p++;
2463                 tok = TOK_A_SHL;
2464             } else {
2465                 tok = TOK_SHL;
2466             }
2467         } else {
2468             tok = TOK_LT;
2469         }
2470         break;
2471
2472     case '>':
2473         PEEKC(c, p);
2474         if (c == '=') {
2475             p++;
2476             tok = TOK_GE;
2477         } else if (c == '>') {
2478             PEEKC(c, p);
2479             if (c == '=') {
2480                 p++;
2481                 tok = TOK_A_SAR;
2482             } else {
2483                 tok = TOK_SAR;
2484             }
2485         } else {
2486             tok = TOK_GT;
2487         }
2488         break;
2489
2490     case '&':
2491         PEEKC(c, p);
2492         if (c == '&') {
2493             p++;
2494             tok = TOK_LAND;
2495         } else if (c == '=') {
2496             p++;
2497             tok = TOK_A_AND;
2498         } else {
2499             tok = '&';
2500         }
2501         break;
2502
2503     case '|':
2504         PEEKC(c, p);
2505         if (c == '|') {
2506             p++;
2507             tok = TOK_LOR;
2508         } else if (c == '=') {
2509             p++;
2510             tok = TOK_A_OR;
2511         } else {
2512             tok = '|';
2513         }
2514         break;
2515
2516     case '+':
2517         PEEKC(c, p);
2518         if (c == '+') {
2519             p++;
2520             tok = TOK_INC;
2521         } else if (c == '=') {
2522             p++;
2523             tok = TOK_A_ADD;
2524         } else {
2525             tok = '+';
2526         }
2527         break;
2528
2529     case '-':
2530         PEEKC(c, p);
2531         if (c == '-') {
2532             p++;
2533             tok = TOK_DEC;
2534         } else if (c == '=') {
2535             p++;
2536             tok = TOK_A_SUB;
2537         } else if (c == '>') {
2538             p++;
2539             tok = TOK_ARROW;
2540         } else {
2541             tok = '-';
2542         }
2543         break;
2544
2545     PARSE2('!', '!', '=', TOK_NE)
2546     PARSE2('=', '=', '=', TOK_EQ)
2547     PARSE2('*', '*', '=', TOK_A_MUL)
2548     PARSE2('%', '%', '=', TOK_A_MOD)
2549     PARSE2('^', '^', '=', TOK_A_XOR)
2550
2551         /* comments or operator */
2552     case '/':
2553         PEEKC(c, p);
2554         if (c == '*') {
2555             p = parse_comment(p);
2556             /* comments replaced by a blank */
2557             tok = ' ';
2558             goto keep_tok_flags;
2559         } else if (c == '/') {
2560             p = parse_line_comment(p);
2561             tok = ' ';
2562             goto keep_tok_flags;
2563         } else if (c == '=') {
2564             p++;
2565             tok = TOK_A_DIV;
2566         } else {
2567             tok = '/';
2568         }
2569         break;
2570
2571         /* simple tokens */
2572     case '(':
2573     case ')':
2574     case '[':
2575     case ']':
2576     case '{':
2577     case '}':
2578     case ',':
2579     case ';':
2580     case ':':
2581     case '?':
2582     case '~':
2583     case '$': /* only used in assembler */
2584     case '@': /* dito */
2585         tok = c;
2586         p++;
2587         break;
2588     default:
2589         tcc_error("unrecognized character \\x%02x", c);
2590         break;
2591     }
2592     tok_flags = 0;
2593 keep_tok_flags:
2594     file->buf_ptr = p;
2595 #if defined(PARSE_DEBUG)
2596     printf("token = %s\n", get_tok_str(tok, &tokc));
2597 #endif
2598 }
2599
2600 /* return next token without macro substitution. Can read input from
2601    macro_ptr buffer */
2602 static void next_nomacro_spc(void)
2603 {
2604     if (macro_ptr) {
2605     redo:
2606         tok = *macro_ptr;
2607         if (tok) {
2608             TOK_GET(&tok, &macro_ptr, &tokc);
2609             if (tok == TOK_LINENUM) {
2610                 file->line_num = tokc.i;
2611                 goto redo;
2612             }
2613         }
2614     } else {
2615         next_nomacro1();
2616     }
2617 }
2618
2619 ST_FUNC void next_nomacro(void)
2620 {
2621     do {
2622         next_nomacro_spc();
2623     } while (is_space(tok));
2624 }
2625
2626 /* substitute arguments in replacement lists in macro_str by the values in
2627    args (field d) and return allocated string */
2628 static int *macro_arg_subst(Sym **nested_list, const int *macro_str, Sym *args)
2629 {
2630     int last_tok, t, spc;
2631     const int *st;
2632     Sym *s;
2633     CValue cval;
2634     TokenString str;
2635     CString cstr;
2636
2637     tok_str_new(&str);
2638     last_tok = 0;
2639     while(1) {
2640         TOK_GET(&t, &macro_str, &cval);
2641         if (!t)
2642             break;
2643         if (t == '#') {
2644             /* stringize */
2645             TOK_GET(&t, &macro_str, &cval);
2646             if (!t)
2647                 break;
2648             s = sym_find2(args, t);
2649             if (s) {
2650                 cstr_new(&cstr);
2651                 st = s->d;
2652                 spc = 0;
2653                 while (*st) {
2654                     TOK_GET(&t, &st, &cval);
2655                     if (!check_space(t, &spc))
2656                         cstr_cat(&cstr, get_tok_str(t, &cval));
2657                 }
2658                 cstr.size -= spc;
2659                 cstr_ccat(&cstr, '\0');
2660 #ifdef PP_DEBUG
2661                 printf("stringize: %s\n", (char *)cstr.data);
2662 #endif
2663                 /* add string */
2664                 cval.cstr = &cstr;
2665                 tok_str_add2(&str, TOK_STR, &cval);
2666                 cstr_free(&cstr);
2667             } else {
2668                 tok_str_add2(&str, t, &cval);
2669             }
2670         } else if (t >= TOK_IDENT) {
2671             s = sym_find2(args, t);
2672             if (s) {
2673                 st = s->d;
2674                 /* if '##' is present before or after, no arg substitution */
2675                 if (*macro_str == TOK_TWOSHARPS || last_tok == TOK_TWOSHARPS) {
2676                     /* special case for var arg macros : ## eats the
2677                        ',' if empty VA_ARGS variable. */
2678                     /* XXX: test of the ',' is not 100%
2679                        reliable. should fix it to avoid security
2680                        problems */
2681                     if (gnu_ext && s->type.t &&
2682                         last_tok == TOK_TWOSHARPS &&
2683                         str.len >= 2 && str.str[str.len - 2] == ',') {
2684                         if (*st == TOK_PLCHLDR) {
2685                             /* suppress ',' '##' */
2686                             str.len -= 2;
2687                         } else {
2688                             /* suppress '##' and add variable */
2689                             str.len--;
2690                             goto add_var;
2691                         }
2692                     } else {
2693                         int t1;
2694                     add_var:
2695                         for(;;) {
2696                             TOK_GET(&t1, &st, &cval);
2697                             if (!t1)
2698                                 break;
2699                             tok_str_add2(&str, t1, &cval);
2700                         }
2701                     }
2702                 } else if (*st != TOK_PLCHLDR) {
2703                     /* NOTE: the stream cannot be read when macro
2704                        substituing an argument */
2705                     macro_subst(&str, nested_list, st, NULL);
2706                 }
2707             } else {
2708                 tok_str_add(&str, t);
2709             }
2710         } else {
2711             tok_str_add2(&str, t, &cval);
2712         }
2713         last_tok = t;
2714     }
2715     tok_str_add(&str, 0);
2716     return str.str;
2717 }
2718
2719 static char const ab_month_name[12][4] =
2720 {
2721     "Jan", "Feb", "Mar", "Apr", "May", "Jun",
2722     "Jul", "Aug", "Sep", "Oct", "Nov", "Dec"
2723 };
2724
2725 /* do macro substitution of current token with macro 's' and add
2726    result to (tok_str,tok_len). 'nested_list' is the list of all
2727    macros we got inside to avoid recursing. Return non zero if no
2728    substitution needs to be done */
2729 static int macro_subst_tok(TokenString *tok_str,
2730                            Sym **nested_list, Sym *s, struct macro_level **can_read_stream)
2731 {
2732     Sym *args, *sa, *sa1;
2733     int mstr_allocated, parlevel, *mstr, t, t1, spc;
2734     const int *p;
2735     TokenString str;
2736     char *cstrval;
2737     CValue cval;
2738     CString cstr;
2739     char buf[32];
2740
2741     /* if symbol is a macro, prepare substitution */
2742     /* special macros */
2743     if (tok == TOK___LINE__) {
2744         snprintf(buf, sizeof(buf), "%d", file->line_num);
2745         cstrval = buf;
2746         t1 = TOK_PPNUM;
2747         goto add_cstr1;
2748     } else if (tok == TOK___FILE__) {
2749         cstrval = file->filename;
2750         goto add_cstr;
2751     } else if (tok == TOK___DATE__ || tok == TOK___TIME__) {
2752         time_t ti;
2753         struct tm *tm;
2754
2755         time(&ti);
2756         tm = localtime(&ti);
2757         if (tok == TOK___DATE__) {
2758             snprintf(buf, sizeof(buf), "%s %2d %d",
2759                      ab_month_name[tm->tm_mon], tm->tm_mday, tm->tm_year + 1900);
2760         } else {
2761             snprintf(buf, sizeof(buf), "%02d:%02d:%02d",
2762                      tm->tm_hour, tm->tm_min, tm->tm_sec);
2763         }
2764         cstrval = buf;
2765     add_cstr:
2766         t1 = TOK_STR;
2767     add_cstr1:
2768         cstr_new(&cstr);
2769         cstr_cat(&cstr, cstrval);
2770         cstr_ccat(&cstr, '\0');
2771         cval.cstr = &cstr;
2772         tok_str_add2(tok_str, t1, &cval);
2773         cstr_free(&cstr);
2774     } else {
2775         mstr = s->d;
2776         mstr_allocated = 0;
2777         if (s->type.t == MACRO_FUNC) {
2778             /* NOTE: we do not use next_nomacro to avoid eating the
2779                next token. XXX: find better solution */
2780         redo:
2781             if (macro_ptr) {
2782                 p = macro_ptr;
2783                 while (is_space(t = *p) || TOK_LINEFEED == t)
2784                     ++p;
2785                 if (t == 0 && can_read_stream) {
2786                     /* end of macro stream: we must look at the token
2787                        after in the file */
2788                     struct macro_level *ml = *can_read_stream;
2789                     macro_ptr = NULL;
2790                     if (ml)
2791                     {
2792                         macro_ptr = ml->p;
2793                         ml->p = NULL;
2794                         *can_read_stream = ml -> prev;
2795                     }
2796                     /* also, end of scope for nested defined symbol */
2797                     (*nested_list)->v = -1;
2798                     goto redo;
2799                 }
2800             } else {
2801                 ch = file->buf_ptr[0];
2802                 while (is_space(ch) || ch == '\n' || ch == '/')
2803                   {
2804                     if (ch == '/')
2805                       {
2806                         int c;
2807                         uint8_t *p = file->buf_ptr;
2808                         PEEKC(c, p);
2809                         if (c == '*') {
2810                             p = parse_comment(p);
2811                             file->buf_ptr = p - 1;
2812                         } else if (c == '/') {
2813                             p = parse_line_comment(p);
2814                             file->buf_ptr = p - 1;
2815                         } else
2816                           break;
2817                       }
2818                     cinp();
2819                   }
2820                 t = ch;
2821             }
2822             if (t != '(') /* no macro subst */
2823                 return -1;
2824
2825             /* argument macro */
2826             next_nomacro();
2827             next_nomacro();
2828             args = NULL;
2829             sa = s->next;
2830             /* NOTE: empty args are allowed, except if no args */
2831             for(;;) {
2832                 /* handle '()' case */
2833                 if (!args && !sa && tok == ')')
2834                     break;
2835                 if (!sa)
2836                     tcc_error("macro '%s' used with too many args",
2837                           get_tok_str(s->v, 0));
2838                 tok_str_new(&str);
2839                 parlevel = spc = 0;
2840                 /* NOTE: non zero sa->t indicates VA_ARGS */
2841                 while ((parlevel > 0 ||
2842                         (tok != ')' &&
2843                          (tok != ',' || sa->type.t))) &&
2844                        tok != -1) {
2845                     if (tok == '(')
2846                         parlevel++;
2847                     else if (tok == ')')
2848                         parlevel--;
2849                     if (tok == TOK_LINEFEED)
2850                         tok = ' ';
2851                     if (!check_space(tok, &spc))
2852                         tok_str_add2(&str, tok, &tokc);
2853                     next_nomacro_spc();
2854                 }
2855                 if (!str.len)
2856                     tok_str_add(&str, TOK_PLCHLDR);
2857                 str.len -= spc;
2858                 tok_str_add(&str, 0);
2859                 sa1 = sym_push2(&args, sa->v & ~SYM_FIELD, sa->type.t, 0);
2860                 sa1->d = str.str;
2861                 sa = sa->next;
2862                 if (tok == ')') {
2863                     /* special case for gcc var args: add an empty
2864                        var arg argument if it is omitted */
2865                     if (sa && sa->type.t && gnu_ext)
2866                         continue;
2867                     else
2868                         break;
2869                 }
2870                 if (tok != ',')
2871                     expect(",");
2872                 next_nomacro();
2873             }
2874             if (sa) {
2875                 tcc_error("macro '%s' used with too few args",
2876                       get_tok_str(s->v, 0));
2877             }
2878
2879             /* now subst each arg */
2880             mstr = macro_arg_subst(nested_list, mstr, args);
2881             /* free memory */
2882             sa = args;
2883             while (sa) {
2884                 sa1 = sa->prev;
2885                 tok_str_free(sa->d);
2886                 sym_free(sa);
2887                 sa = sa1;
2888             }
2889             mstr_allocated = 1;
2890         }
2891         sym_push2(nested_list, s->v, 0, 0);
2892         macro_subst(tok_str, nested_list, mstr, can_read_stream);
2893         /* pop nested defined symbol */
2894         sa1 = *nested_list;
2895         *nested_list = sa1->prev;
2896         sym_free(sa1);
2897         if (mstr_allocated)
2898             tok_str_free(mstr);
2899     }
2900     return 0;
2901 }
2902
2903 /* handle the '##' operator. Return NULL if no '##' seen. Otherwise
2904    return the resulting string (which must be freed). */
2905 static inline int *macro_twosharps(const int *macro_str)
2906 {
2907     const int *ptr;
2908     int t;
2909     TokenString macro_str1;
2910     CString cstr;
2911     int n, start_of_nosubsts;
2912
2913     /* we search the first '##' */
2914     for(ptr = macro_str;;) {
2915         CValue cval;
2916         TOK_GET(&t, &ptr, &cval);
2917         if (t == TOK_TWOSHARPS)
2918             break;
2919         /* nothing more to do if end of string */
2920         if (t == 0)
2921             return NULL;
2922     }
2923
2924     /* we saw '##', so we need more processing to handle it */
2925     start_of_nosubsts = -1;
2926     tok_str_new(&macro_str1);
2927     for(ptr = macro_str;;) {
2928         TOK_GET(&tok, &ptr, &tokc);
2929         if (tok == 0)
2930             break;
2931         if (tok == TOK_TWOSHARPS)
2932             continue;
2933         if (tok == TOK_NOSUBST && start_of_nosubsts < 0)
2934             start_of_nosubsts = macro_str1.len;
2935         while (*ptr == TOK_TWOSHARPS) {
2936             /* given 'a##b', remove nosubsts preceding 'a' */
2937             if (start_of_nosubsts >= 0)
2938                 macro_str1.len = start_of_nosubsts;
2939             /* given 'a##b', skip '##' */
2940             t = *++ptr;
2941             /* given 'a##b', remove nosubsts preceding 'b' */
2942             while (t == TOK_NOSUBST)
2943                 t = *++ptr;
2944             if (t && t != TOK_TWOSHARPS) {
2945                 CValue cval;
2946                 TOK_GET(&t, &ptr, &cval);
2947                 /* We concatenate the two tokens */
2948                 cstr_new(&cstr);
2949                 if (tok != TOK_PLCHLDR)
2950                     cstr_cat(&cstr, get_tok_str(tok, &tokc));
2951                 n = cstr.size;
2952                 if (t != TOK_PLCHLDR || tok == TOK_PLCHLDR)
2953                     cstr_cat(&cstr, get_tok_str(t, &cval));
2954                 cstr_ccat(&cstr, '\0');
2955
2956                 tcc_open_bf(tcc_state, ":paste:", cstr.size);
2957                 memcpy(file->buffer, cstr.data, cstr.size);
2958                 for (;;) {
2959                     next_nomacro1();
2960                     if (0 == *file->buf_ptr)
2961                         break;
2962                     tok_str_add2(&macro_str1, tok, &tokc);
2963                     tcc_warning("pasting \"%.*s\" and \"%s\" does not give a valid preprocessing token",
2964                         n, cstr.data, (char*)cstr.data + n);
2965                 }
2966                 tcc_close();
2967                 cstr_free(&cstr);
2968             }
2969         }
2970         if (tok != TOK_NOSUBST) {
2971             tok_str_add2(&macro_str1, tok, &tokc);
2972             tok = ' ';
2973             start_of_nosubsts = -1;
2974         }
2975         tok_str_add2(&macro_str1, tok, &tokc);
2976     }
2977     tok_str_add(&macro_str1, 0);
2978     return macro_str1.str;
2979 }
2980
2981
2982 /* do macro substitution of macro_str and add result to
2983    (tok_str,tok_len). 'nested_list' is the list of all macros we got
2984    inside to avoid recursing. */
2985 static void macro_subst(TokenString *tok_str, Sym **nested_list,
2986                         const int *macro_str, struct macro_level ** can_read_stream)
2987 {
2988     Sym *s;
2989     int *macro_str1;
2990     const int *ptr;
2991     int t, ret, spc;
2992     CValue cval;
2993     struct macro_level ml;
2994     int force_blank;
2995
2996     /* first scan for '##' operator handling */
2997     ptr = macro_str;
2998     macro_str1 = macro_twosharps(ptr);
2999
3000     if (macro_str1)
3001         ptr = macro_str1;
3002     spc = 0;
3003     force_blank = 0;
3004
3005     while (1) {
3006         /* NOTE: ptr == NULL can only happen if tokens are read from
3007            file stream due to a macro function call */
3008         if (ptr == NULL)
3009             break;
3010         TOK_GET(&t, &ptr, &cval);
3011         if (t == 0)
3012             break;
3013         if (t == TOK_NOSUBST) {
3014             /* following token has already been subst'd. just copy it on */
3015             tok_str_add2(tok_str, TOK_NOSUBST, NULL);
3016             TOK_GET(&t, &ptr, &cval);
3017             goto no_subst;
3018         }
3019         s = define_find(t);
3020         if (s != NULL) {
3021             /* if nested substitution, do nothing */
3022             if (sym_find2(*nested_list, t)) {
3023                 /* and mark it as TOK_NOSUBST, so it doesn't get subst'd again */
3024                 tok_str_add2(tok_str, TOK_NOSUBST, NULL);
3025                 goto no_subst;
3026             }
3027             ml.p = macro_ptr;
3028             if (can_read_stream)
3029                 ml.prev = *can_read_stream, *can_read_stream = &ml;
3030             macro_ptr = (int *)ptr;
3031             tok = t;
3032             ret = macro_subst_tok(tok_str, nested_list, s, can_read_stream);
3033             ptr = (int *)macro_ptr;
3034             macro_ptr = ml.p;
3035             if (can_read_stream && *can_read_stream == &ml)
3036                 *can_read_stream = ml.prev;
3037             if (ret != 0)
3038                 goto no_subst;
3039             if (parse_flags & PARSE_FLAG_SPACES)
3040                 force_blank = 1;
3041         } else {
3042         no_subst:
3043             if (force_blank) {
3044                 tok_str_add(tok_str, ' ');
3045                 spc = 1;
3046                 force_blank = 0;
3047             }
3048             if (!check_space(t, &spc))
3049                 tok_str_add2(tok_str, t, &cval);
3050         }
3051     }
3052     if (macro_str1)
3053         tok_str_free(macro_str1);
3054 }
3055
3056 /* return next token with macro substitution */
3057 ST_FUNC void next(void)
3058 {
3059     Sym *nested_list, *s;
3060     TokenString str;
3061     struct macro_level *ml;
3062
3063  redo:
3064     if (parse_flags & PARSE_FLAG_SPACES)
3065         next_nomacro_spc();
3066     else
3067         next_nomacro();
3068     if (!macro_ptr) {
3069         /* if not reading from macro substituted string, then try
3070            to substitute macros */
3071         if (tok >= TOK_IDENT &&
3072             (parse_flags & PARSE_FLAG_PREPROCESS)) {
3073             s = define_find(tok);
3074             if (s) {
3075                 /* we have a macro: we try to substitute */
3076                 tok_str_new(&str);
3077                 nested_list = NULL;
3078                 ml = NULL;
3079                 if (macro_subst_tok(&str, &nested_list, s, &ml) == 0) {
3080                     /* substitution done, NOTE: maybe empty */
3081                     tok_str_add(&str, 0);
3082                     macro_ptr = str.str;
3083                     macro_ptr_allocated = str.str;
3084                     goto redo;
3085                 }
3086             }
3087         }
3088     } else {
3089         if (tok == 0) {
3090             /* end of macro or end of unget buffer */
3091             if (unget_buffer_enabled) {
3092                 macro_ptr = unget_saved_macro_ptr;
3093                 unget_buffer_enabled = 0;
3094             } else {
3095                 /* end of macro string: free it */
3096                 tok_str_free(macro_ptr_allocated);
3097                 macro_ptr_allocated = NULL;
3098                 macro_ptr = NULL;
3099             }
3100             goto redo;
3101         } else if (tok == TOK_NOSUBST) {
3102             /* discard preprocessor's nosubst markers */
3103             goto redo;
3104         }
3105     }
3106
3107     /* convert preprocessor tokens into C tokens */
3108     if (tok == TOK_PPNUM &&
3109         (parse_flags & PARSE_FLAG_TOK_NUM)) {
3110         parse_number((char *)tokc.cstr->data);
3111     }
3112 }
3113
3114 /* push back current token and set current token to 'last_tok'. Only
3115    identifier case handled for labels. */
3116 ST_INLN void unget_tok(int last_tok)
3117 {
3118     int i, n;
3119     int *q;
3120     if (unget_buffer_enabled)
3121       {
3122         /* assert(macro_ptr == unget_saved_buffer + 1);
3123            assert(*macro_ptr == 0);  */
3124       }
3125     else
3126       {
3127         unget_saved_macro_ptr = macro_ptr;
3128         unget_buffer_enabled = 1;
3129       }
3130     q = unget_saved_buffer;
3131     macro_ptr = q;
3132     *q++ = tok;
3133     n = tok_ext_size(tok) - 1;
3134     for(i=0;i<n;i++)
3135         *q++ = tokc.tab[i];
3136     *q = 0; /* end of token string */
3137     tok = last_tok;
3138 }
3139
3140
3141 /* better than nothing, but needs extension to handle '-E' option
3142    correctly too */
3143 ST_FUNC void preprocess_init(TCCState *s1)
3144 {
3145     s1->include_stack_ptr = s1->include_stack;
3146     /* XXX: move that before to avoid having to initialize
3147        file->ifdef_stack_ptr ? */
3148     s1->ifdef_stack_ptr = s1->ifdef_stack;
3149     file->ifdef_stack_ptr = s1->ifdef_stack_ptr;
3150
3151     vtop = vstack - 1;
3152     s1->pack_stack[0] = 0;
3153     s1->pack_stack_ptr = s1->pack_stack;
3154 }
3155
3156 ST_FUNC void preprocess_new(void)
3157 {
3158     int i, c;
3159     const char *p, *r;
3160
3161     /* init isid table */
3162     for(i=CH_EOF;i<256;i++)
3163         isidnum_table[i-CH_EOF] = isid(i) || isnum(i);
3164
3165     /* add all tokens */
3166     table_ident = NULL;
3167     memset(hash_ident, 0, TOK_HASH_SIZE * sizeof(TokenSym *));
3168
3169     tok_ident = TOK_IDENT;
3170     p = tcc_keywords;
3171     while (*p) {
3172         r = p;
3173         for(;;) {
3174             c = *r++;
3175             if (c == '\0')
3176                 break;
3177         }
3178         tok_alloc(p, r - p - 1);
3179         p = r;
3180     }
3181 }
3182
3183 static void line_macro_output(BufferedFile *f, const char *s, TCCState *s1)
3184 {
3185     switch (s1->Pflag) {
3186         case LINE_MACRO_OUTPUT_FORMAT_STD:
3187             /* "tcc -E -P1" case */
3188             fprintf(s1->ppfp, "# line %d \"%s\"\n", f->line_num, f->filename);
3189             break;
3190
3191         case LINE_MACRO_OUTPUT_FORMAT_NONE:
3192             /* "tcc -E -P" case: don't output a line directive */
3193             break;
3194
3195         case LINE_MACRO_OUTPUT_FORMAT_GCC:
3196         default:
3197             /* "tcc -E" case: a gcc standard by default */
3198             fprintf(s1->ppfp, "# %d \"%s\"%s\n", f->line_num, f->filename, s);
3199             break;
3200     }
3201 }
3202
3203 /* Preprocess the current file */
3204 ST_FUNC int tcc_preprocess(TCCState *s1)
3205 {
3206     BufferedFile *file_ref, **iptr, **iptr_new;
3207     int token_seen, d;
3208     const char *s;
3209
3210     preprocess_init(s1);
3211     ch = file->buf_ptr[0];
3212     tok_flags = TOK_FLAG_BOL | TOK_FLAG_BOF;
3213     parse_flags = PARSE_FLAG_ASM_COMMENTS | PARSE_FLAG_PREPROCESS |
3214         PARSE_FLAG_LINEFEED | PARSE_FLAG_SPACES;
3215     token_seen = 0;
3216     file->line_ref = 0;
3217     file_ref = NULL;
3218     iptr = s1->include_stack_ptr;
3219
3220     for (;;) {
3221         next();
3222         if (tok == TOK_EOF) {
3223             break;
3224         } else if (file != file_ref) {
3225             if (file_ref)
3226                 line_macro_output(file_ref, "", s1);
3227             goto print_line;
3228         } else if (tok == TOK_LINEFEED) {
3229             if (!token_seen)
3230                 continue;
3231             file->line_ref++;
3232             token_seen = 0;
3233         } else if (!token_seen) {
3234             d = file->line_num - file->line_ref;
3235             if (file != file_ref || d >= 8) {
3236 print_line:
3237                 s = "";
3238                 if (tcc_state->Pflag == LINE_MACRO_OUTPUT_FORMAT_GCC) {
3239                     iptr_new = s1->include_stack_ptr;
3240                     s = iptr_new > iptr ? " 1"
3241                     : iptr_new < iptr ? " 2"
3242                     : iptr_new > s1->include_stack ? " 3"
3243                     : ""
3244                     ;
3245                 }
3246                 line_macro_output(file, s, s1);
3247             } else {
3248                 while (d > 0)
3249                     fputs("\n", s1->ppfp), --d;
3250             }
3251             file->line_ref = (file_ref = file)->line_num;
3252             token_seen = tok != TOK_LINEFEED;
3253             if (!token_seen)
3254                 continue;
3255         }
3256         fputs(get_tok_str(tok, &tokc), s1->ppfp);
3257     }
3258     return 0;
3259 }